threads
listlengths 1
2.99k
|
|---|
[
{
"msg_contents": "\nIs the CVS server down at the moment?\n\nI've just tried to get a latest source set down, and are getting:\n\n[peter@maidast pgsql]$ cvs update\nFatal error, aborting.\n: no such user\n\nAlso, lately it's been taking hours (literally) for a fresh update to\noccur, with the connection timing out if I don't keep an eye on it.\n\n-- \n Peter T Mount [email protected]\n Main Homepage: http://www.retep.org.uk\nPostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n Java PDF Generator: http://www.retep.org.uk/pdf\n\n",
"msg_date": "Sun, 18 Oct 1998 09:55:12 +0100 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": true,
"msg_subject": "CVS Server down?"
}
] |
[
{
"msg_contents": "I got error message trying connect to cvs.\n\nFatal error, aborting.\n: no such user\n\n\tOleg\n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n",
"msg_date": "Sun, 18 Oct 1998 14:07:06 +0400 (MSD)",
"msg_from": "Oleg Bartunov <[email protected]>",
"msg_from_op": true,
"msg_subject": "problem with cvs ?"
}
] |
[
{
"msg_contents": "Thus spake Paul A Vixie\n> if someone wants to be able to represent a host address and a netmask,\n> then may i suggest that they use the same type, except they'll have to\n> use two of them, one for the host address and one for the netmask. in\n> this case the prefix length would like to be enforced/implied at /32,\n> and never printed. this is not a CIDR. it's how BIND's \"inet_ntop\"\n> and \"inet_pton\" work.\n\nIf you are forcing/implying the netmask in the host type, why not just use\nit for the netmask? The point of my functions is to extract the different\nparts from that.\n\n> if someone wants to be able to represent a CIDRized host address, that\n> is, a host and prefix, which means the host-part \"can be\" zero or non-zero,\n> and the prefix might or might not be required in input or printed in output,\n> then that's an entirely different thing from either of the above. and\n> while i've got the code almost ready to do that, i don't want this behaviour\n> in my own target application for the original \"cidr\" prototype i sent in.\n> \n> it looks to me as though we've got three new pgsql types here. you agree?\n\nAs far as I understand, that last is the only one we were expecting in\nPostgreSQL. So how about this? Let's put back all of the original inet\nstuff but call it cidr. Finish your stuff and we'll put it in as the\ninet type. Personally I think we just need the latter but I would\nrather have both than neither.\n\nI just noticed that this wasn't copied to the list. I hope you don't\nmind me copying this response there so that others can add their\nopinions too.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Sun, 18 Oct 1998 06:15:15 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> If you are forcing/implying the netmask in the host type, why not just use\n> it for the netmask? The point of my functions is to extract the different\n> parts from that.\n\nFunctions to extract it is one thing. What the data type can contain is\nanother. I'm building a registry database of CIDRized IP delegations and\nI need it to be an error to insert an element whose host-part is nonzero.\nThat was the launch application for the original \"cidr\" type I contributed,\nand in the resulting hullabaloo we ended up trying to represent \"hosts\"\nwhich had potential nonzero content across all 32 bits even though the\nnetmask was less than 32 bits long. I can't make any good use of that in\nmy application.\n\nAgain, the question of whether there are functions available to build up\nor tear apart these compound values is not relevant to my point here. It's\na good idea to have such functions if the type is constructed as has been\nproposed here during the recent hullabaloo.\n\nBut in my application, I need to be able to key a table to a network block,\nand it is literally meaningless to have a nonzero host part or a fixed size\nmantissa in that application.\n\nSince other people want to use this type differently, I propose that we make\nit into two types: INET for host addresses and CIDR for network addresses.\nBoth can include /## to set a netmask if folks really don't like making the\naddress and netmask into two adjacent columns. But CIDR has to disallow\nnonzero host-parts or it doesn't mean what I need it to mean.\n\nPostgreSQL has an opportunity here to be the first SQL system to be able to\nnatively support IP registries. There's some market share lurking in this.\n\n> > if someone wants to be able to represent a CIDRized host address, that\n> > is, a host and prefix, which means the host-part \"can be\" zero or non-zero,\n> > and the prefix might or might not be required in input or printed in output\n> > then that's an entirely different thing from either of the above. and while\n> > i've got the code almost ready to do that, i don't want this behaviour\n> > in my own target application for the original \"cidr\" prototype i sent in.\n> > \n> > it looks to me as though we've got three new pgsql types here. you agree?\n> \n> As far as I understand, that last is the only one we were expecting in\n> PostgreSQL. So how about this? Let's put back all of the original inet\n> stuff but call it cidr. Finish your stuff and we'll put it in as the\n> inet type. Personally I think we just need the latter but I would\n> rather have both than neither.\n\nOk. Will you take care of integrating both types, including indexing? (My\ninability to use the original cidr type for an index was my original problem.)\n\n> I just noticed that this wasn't copied to the list. I hope you don't\n> mind me copying this response there so that others can add their\n> opinions too.\n\nActually I do mind, and had I been addressing the larger audience my note\nwould have been a lot clearer. But now that we've moved the discussion here\nI'm CC'ing the list on my response as well. No harm done.\n",
"msg_date": "Sun, 18 Oct 1998 11:22:32 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "Thus spake Paul A Vixie\n> Since other people want to use this type differently, I propose that we make\n> it into two types: INET for host addresses and CIDR for network addresses.\n> Both can include /## to set a netmask if folks really don't like making the\n> address and netmask into two adjacent columns. But CIDR has to disallow\n> nonzero host-parts or it doesn't mean what I need it to mean.\n\nI am cool with this. In fact I think I mentioned earlier in the thread that\nthis requirement (it was a theoretical possibility to me at the time) was\nthe one reason why two different types might be required. OK so that\nmeans that, as I suggested, we need two underlying function sets, the\ninet_net_* ones and the inet_cidr_* ones (might be some efficiency\npossible if they are in the same source file and use some common\nroutines) and the original inet.c file gets changed to cidr.c and\nturned into the cidr type. I'll take the existing inet.c as it exists\non my system and submit it as the _new_ inet type.\n\nOf course we can both use the same structure to store the data. I suggest\nwe leave the name as inet since cidr seems to be a special case of inet.\nDo you agree?\n\n> Ok. Will you take care of integrating both types, including indexing? (My\n> inability to use the original cidr type for an index was my original problem.)\n\nIf you mean adding to the catalogues I believe Bruce is dealing with that.\nI will make a new cidr.c file which is basically the original one with\na few replacements for the new names. Will you have the final inet_net_*\nand inet_cidr_* functions in the tree shortly?\n\n> > I just noticed that this wasn't copied to the list. I hope you don't\n> > mind me copying this response there so that others can add their\n> > opinions too.\n> \n> Actually I do mind, and had I been addressing the larger audience my note\n> would have been a lot clearer. But now that we've moved the discussion here\n> I'm CC'ing the list on my response as well. No harm done.\n\nOops. Sorry about that. I thought that you simply forgot to copy\nthe list. Anyway, there are a few people with fingers in this\nparticular pie so it's probably just as well. This is the hackers\nlist so I don't imagine many people found your message unclear. I\ncertainly didn't.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Sun, 18 Oct 1998 23:29:08 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> Thus spake Paul A Vixie\n> > Since other people want to use this type differently, I propose that we make\n> > it into two types: INET for host addresses and CIDR for network addresses.\n> > Both can include /## to set a netmask if folks really don't like making the\n> > address and netmask into two adjacent columns. But CIDR has to disallow\n> > nonzero host-parts or it doesn't mean what I need it to mean.\n> \n> I am cool with this. In fact I think I mentioned earlier in the thread that\n> this requirement (it was a theoretical possibility to me at the time) was\n> the one reason why two different types might be required. OK so that\n> means that, as I suggested, we need two underlying function sets, the\n> inet_net_* ones and the inet_cidr_* ones (might be some efficiency\n> possible if they are in the same source file and use some common\n> routines) and the original inet.c file gets changed to cidr.c and\n> turned into the cidr type. I'll take the existing inet.c as it exists\n> on my system and submit it as the _new_ inet type.\n\nSorry, but I just have to ask. I am not sure what the issues are, but\nit is possible to have them all be the same type. Have a netmask field\ninside the type and a cidr field inside the type, and just use one of\nthem at a time for any given entry? Only one byte each, right? I sure\nwould like to avoid \"type/function bloat\". Also, could a RULE be\ncreated to simulate the network restriction Paul requires?\n\nMaybe it is clearer to have two types, with different purposes. I am\njust asking, and if people are going to need functions to convert\nbetween the two types, it may be worth merging them. I was thinking you\ncould display them differently based on which field they used.\n\nThis is all just a guess.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 01:47:39 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> > Thus spake Paul A Vixie\n> > > Since other people want to use this type differently, I propose that we make\n> > > it into two types: INET for host addresses and CIDR for network addresses.\n> > > Both can include /## to set a netmask if folks really don't like making the\n> > > address and netmask into two adjacent columns. But CIDR has to disallow\n> > > nonzero host-parts or it doesn't mean what I need it to mean.\n\nPaul, can you give the unknowledgable among us an example of what you\nwant and don't want? Thanks.\n\nI was assuming the type would allow these as hosts:\n\n\t192.4.13.42\n\t192.4.13.42 255.255.255.0\n\t192.4.13.42/32\n\t192.4.13.42/32 255.255.255.0\n\nand these are networks:\n\n\t192.63.91.234/28\n\t192.63.91.234/28 255.255.255.128\n\nAny non-32 '/' is a network? Can a network have a netmask? Not sure.\nInternally, a -1 in a field indicates the user did not supply a cidr\nvalue(assume it is a host), or did not specify a netmask. I thought\nthis is where D'Arcy was going with this.\n\nI assume we are _not_ going to use the /24 notation to specify a\nnetmask, because the cidr '/' notation has never been used for that,\nright?\n\t\n> > \n> > I am cool with this. In fact I think I mentioned earlier in the thread that\n> > this requirement (it was a theoretical possibility to me at the time) was\n> > the one reason why two different types might be required. OK so that\n> > means that, as I suggested, we need two underlying function sets, the\n> > inet_net_* ones and the inet_cidr_* ones (might be some efficiency\n> > possible if they are in the same source file and use some common\n> > routines) and the original inet.c file gets changed to cidr.c and\n> > turned into the cidr type. I'll take the existing inet.c as it exists\n> > on my system and submit it as the _new_ inet type.\n> \n> Sorry, but I just have to ask. I am not sure what the issues are, but\n> it is possible to have them all be the same type. Have a netmask field\n> inside the type and a cidr field inside the type, and just use one of\n> them at a time for any given entry? Only one byte each, right? I sure\n> would like to avoid \"type/function bloat\". Also, could a RULE be\n> created to simulate the network restriction Paul requires?\n> \n> Maybe it is clearer to have two types, with different purposes. I am\n> just asking, and if people are going to need functions to convert\n> between the two types, it may be worth merging them. I was thinking you\n> could display them differently based on which field they used.\n> \n> This is all just a guess.\n\nIf I am wrong about the above, I have one more question. Would an\natttypmod setting for each column help? What about a compile-time\ndefine?\n\nI know Paul is a big name, but are the duplicate types meaningful for\nordinary users, or would they prefer just one type. If they would\nprefer one type, we can do that, and make sure Paul gets what he wants\ntoo.\n\nWant something really fancy?\n\n\tCREATE TABLE hostnet(\n\t\thost \tinet(host),\n\t\tnetwork\tinet(net),\n\t\tmisc\tinet\n\t);\n\nThis is possible.\n\nIf we can get some more rapid-fire e-mails going around, I think we can\nresolve this in the next day or two. We have testing and documenation\nto do, and Marc at some point may pull the plug on us.\n\nI am ready to install whatever you folks come up with.\n\nIf we can come up with one type, that is less bloat/work for me, but my\nprimary goal is that we come up with a solid type/types that we aren't\ngoing to need to redesign in the future, causing problems for existing\nusers.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 06:52:01 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> Sorry, but I just have to ask. I am not sure what the issues are, but\n> it is possible to have them all be the same type. Have a netmask field\n> inside the type and a cidr field inside the type, and just use one of\n> them at a time for any given entry? Only one byte each, right? I sure\n> would like to avoid \"type/function bloat\". Also, could a RULE be\n> created to simulate the network restriction Paul requires?\n> \n> Maybe it is clearer to have two types, with different purposes. I am\n> just asking, and if people are going to need functions to convert\n> between the two types, it may be worth merging them. I was thinking you\n> could display them differently based on which field they used.\n\nI can think of two possibilities. I thought of this before for other\npurposes but I never asked about it. Can we set up inet_in to take\na second argument? If so then we can set up two types but set them\nup such that one calls the in function with 0 and one with 1. We\nwould store that value in a new field in the structure and check\nthat for every other function where necessary.\n\nIf not then we can do the same thing by creating two wrapper input\nfunctions which call the one I have described with the extra argument.\nAlmost as good.\n\nIf this makes sense I suggest we go forward with our existing plan and\nlook to fold it after 6.4. It's minor bloat for now and it would\nsimplify getting everything in there.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Mon, 19 Oct 1998 07:59:21 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> Paul, can you give the unknowledgable among us an example of what you\n> want and don't want? Thanks.\n\nLet me take a stab at it too.\n\n> I was assuming the type would allow these as hosts:\n> \n> \t192.4.13.42\n> \t192.4.13.42 255.255.255.0\n\nAn alternate syntax of 192.4.13.42:255.255.255.0 was proposed but won't\nbe implemented, at least at this time. Maybe put it into TODO for 6.4++.\nIn the meantime, this should be 192.4.13.42/24.\n\n> \t192.4.13.42/32\n> \t192.4.13.42/32 255.255.255.0\n\nBasically two different netmasks. I think you mean 192.4.13.42/24 here too.\n\n> and these are networks:\n> \n> \t192.63.91.234/28\n\nThis is a host in a 16 IP subnet. As a cidr type (as proposed) this\nwould be invalid. Note the 234 is 11101010b and the /28 implies a\ntrailing octet of 11110000b. A valid entry would have been something\nlike 192.63.91.224/28.\n\n> \t192.63.91.234/28 255.255.255.128\n\nAgain, the /28 and the 255.255.255.128 are trying to specify the same\nthing but they are different. The /28 is an alternate way of saying\na netmask of 255.255.255.240. 255.255.255.128 would be /25. Either\nway, this is invalid as a cidr but valid as an inet host on a network.\n\n> Any non-32 '/' is a network? Can a network have a netmask? Not sure.\n\nNope. /32 is a valid network. It's a class D in the old class system.\nThat's why I suggested -1 for an unspecified netmask to differentiate\nit from a network of one host.\n\nA network can certainly have a netmask. In fact, it always does. It\ncan be implied in certain cases. The following (if I understand Paul's\nproposal - correct me if not) show the relationships.\n\nInput cidr output inet output\n============= ================ ================\n192.63.91.234 192.63.91.234/32\t192.63.91.234/32\n192.63.91 192.63.91/24 192.63.91.0/24\n192.63 192.63/16 192.63.0.0/16\n192 192/8 192.0.0.0/8\n\nThis look right to you, Paul?\n\n> Internally, a -1 in a field indicates the user did not supply a cidr\n> value(assume it is a host), or did not specify a netmask. I thought\n> this is where D'Arcy was going with this.\n\nClose. A -1 is only valid for the inet type, not the cidr type. This\nis because the /XX effectively specifies both the network and netmask.\nUnder cidr this must be specified. Under inet it may be. If it isn't,\nor -1 is specified, -1 is stored as the bits and it is not output on\ndisplay. This allows it to be used as a host or as a host/network\ncombination. Note that there is nothing stopping you from using the\ninet type to store networks too. It's just that using cidr does the\nextra error checking and outputs in the more standard network format.\n\n> I assume we are _not_ going to use the /24 notation to specify a\n> netmask, because the cidr '/' notation has never been used for that,\n> right?\n\nWell, it can be implied but I think it is OK to specify. 192.63.91 and\n192.63.91/24 should be the same thing. If the number of bits is not\n32, 24, 16 or 8 you have to specify it.\n\n> If I am wrong about the above, I have one more question. Would an\n> atttypmod setting for each column help? What about a compile-time\n> define?\n\nWe discussed this at one point. I think that is more useful for\nspecifying output formats. For example, 192.63.91.234/24 is identical\nto 192.63.91.234:255.255.255.0 (if we add that format) but I think\nthat's 6.4++ too. I think it would also only apply to the inet type\nbut Paul should know.\n\n> I know Paul is a big name, but are the duplicate types meaningful for\n> ordinary users, or would they prefer just one type. If they would\n> prefer one type, we can do that, and make sure Paul gets what he wants\n> too.\n\nI understand why Paul needs his type but I think the inet type is\nvaluable too. I think my suggestion above is a good compromise.\n\n> If we can get some more rapid-fire e-mails going around, I think we can\n> resolve this in the next day or two. We have testing and documenation\n> to do, and Marc at some point may pull the plug on us.\n\n> If we can come up with one type, that is less bloat/work for me, but my\n> primary goal is that we come up with a solid type/types that we aren't\n> going to need to redesign in the future, causing problems for existing\n> users.\n\nI think we are just about there. If we go with my plan (completely\ndifferent functionality for now and fold it later) there should be\nno API change later. There will be code and catalogue changes but\nthey should be relatively painless.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Mon, 19 Oct 1998 08:44:57 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> A network can certainly have a netmask. In fact, it always does. It\n> can be implied in certain cases. The following (if I understand Paul's\n> proposal - correct me if not) show the relationships.\n> \n> Input cidr output inet output\n> ============= ================ ================\n> 192.63.91.234 192.63.91.234/32\t192.63.91.234/32\n> 192.63.91 192.63.91/24 192.63.91.0/24\n> 192.63 192.63/16 192.63.0.0/16\n> 192 192/8 192.0.0.0/8\n> \n> This look right to you, Paul?\n\nno. the last three inputs are not valid where a host address is expected.\n\n> > If I am wrong about the above, I have one more question. Would an\n> > atttypmod setting for each column help? What about a compile-time\n> > define?\n> \n> We discussed this at one point. I think that is more useful for\n> specifying output formats. For example, 192.63.91.234/24 is identical\n> to 192.63.91.234:255.255.255.0 (if we add that format) but I think\n> that's 6.4++ too. I think it would also only apply to the inet type\n> but Paul should know.\n> \n> > I know Paul is a big name, but are the duplicate types meaningful for\n> > ordinary users, or would they prefer just one type. If they would\n> > prefer one type, we can do that, and make sure Paul gets what he wants\n> > too.\n> \n> I understand why Paul needs his type but I think the inet type is\n> valuable too. I think my suggestion above is a good compromise.\n\nbigness of names doesn't matter. applications matter. i can see a use for\nboth types, but they are inherently different types. a host that has a\nnetmask which can be expressed in cidr notation is one such type. a net\nthat has a netmask which must be expressed in cidr notation is another such\ntype. the difference comes down to \"host part must be zero\" for the network\ntype. there are also some minor differences in the input/output formats,\nsince a host address always has four octets on both input and output, while\na network only prints as many octets as the cidr width specifies, and these\nare the only required octets on input (though extra .0's can be specified).\n\n> I think we are just about there. If we go with my plan (completely\n> different functionality for now and fold it later) there should be\n> no API change later. There will be code and catalogue changes but\n> they should be relatively painless.\n\nso shall i test the inet_cidr_ functions and punt them on in?\n\n",
"msg_date": "Mon, 19 Oct 1998 12:07:30 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "Can't we just use a CONSTRAINT where a host address is expected? That sounds\neasier than setting up two different types to me...\n\nTaral\n\n> -----Original Message-----\n> From: [email protected]\n> [mailto:[email protected]]On Behalf Of Paul A Vixie\n> Sent: Monday, October 19, 1998 2:08 PM\n> To: [email protected]\n> Subject: Re: [HACKERS] Re: inet/cidr/bind\n>\n>\n> > A network can certainly have a netmask. In fact, it always does. It\n> > can be implied in certain cases. The following (if I understand Paul's\n> > proposal - correct me if not) show the relationships.\n> >\n> > Input cidr output inet output\n> > ============= ================ ================\n> > 192.63.91.234 192.63.91.234/32\t192.63.91.234/32\n> > 192.63.91 192.63.91/24 192.63.91.0/24\n> > 192.63 192.63/16 192.63.0.0/16\n> > 192 192/8 192.0.0.0/8\n> >\n> > This look right to you, Paul?\n>\n> no. the last three inputs are not valid where a host address is expected.\n>\n> > > If I am wrong about the above, I have one more question. Would an\n> > > atttypmod setting for each column help? What about a compile-time\n> > > define?\n> >\n> > We discussed this at one point. I think that is more useful for\n> > specifying output formats. For example, 192.63.91.234/24 is identical\n> > to 192.63.91.234:255.255.255.0 (if we add that format) but I think\n> > that's 6.4++ too. I think it would also only apply to the inet type\n> > but Paul should know.\n> >\n> > > I know Paul is a big name, but are the duplicate types meaningful for\n> > > ordinary users, or would they prefer just one type. If they would\n> > > prefer one type, we can do that, and make sure Paul gets what he wants\n> > > too.\n> >\n> > I understand why Paul needs his type but I think the inet type is\n> > valuable too. I think my suggestion above is a good compromise.\n>\n> bigness of names doesn't matter. applications matter. i can see\n> a use for\n> both types, but they are inherently different types. a host that has a\n> netmask which can be expressed in cidr notation is one such type. a net\n> that has a netmask which must be expressed in cidr notation is\n> another such\n> type. the difference comes down to \"host part must be zero\" for\n> the network\n> type. there are also some minor differences in the input/output formats,\n> since a host address always has four octets on both input and\n> output, while\n> a network only prints as many octets as the cidr width specifies,\n> and these\n> are the only required octets on input (though extra .0's can be\n> specified).\n>\n> > I think we are just about there. If we go with my plan (completely\n> > different functionality for now and fold it later) there should be\n> > no API change later. There will be code and catalogue changes but\n> > they should be relatively painless.\n>\n> so shall i test the inet_cidr_ functions and punt them on in?\n>\n>\n\n",
"msg_date": "Mon, 19 Oct 1998 14:43:17 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "Thus spake Paul A Vixie\n> > Input cidr output inet output\n> > ============= ================ ================\n> > 192.63.91.234 192.63.91.234/32\t192.63.91.234/32\n> > 192.63.91 192.63.91/24 192.63.91.0/24\n> > 192.63 192.63/16 192.63.0.0/16\n> > 192 192/8 192.0.0.0/8\n> > \n> > This look right to you, Paul?\n> \n> no. the last three inputs are not valid where a host address is expected.\n\nCan you amplify? Is it correct as far as cidr goes? If so, I have no\nproblem making it an error for the inet type. My thinking was based\non the earlier concept of having one type and accepting networks in it.\nIf we have the separate cidr type then I guess inet should always require\n4 octets (until ipv6 anyway) and cidr should be used for networks.\n\nHow about something like 192.63.0.0/16? Should that be an error under the\ninet type since it is the network? I am thinking not since technically\n192.63.0.0 is a valid host under 192.63/16 although it is generally\navoided since there is still software that assumes that it is the\nnetwork or even the broadcast.\n\n> > I understand why Paul needs his type but I think the inet type is\n> > valuable too. I think my suggestion above is a good compromise.\n> \n> bigness of names doesn't matter. applications matter. i can see a use for\n> both types, but they are inherently different types. a host that has a\n\nI agree although I did change my mind somewhat based on these discussions.\n\nAnd besides:\n\n if (strlen(\"D'Arcy J.M. Cain\") > strlen(\"Paul Vixie\"))\n printf(\"Who's the big name now? :-)\\n\");\n\nAnd look at all that StUdLy punctuation. :-)\n\n> a network only prints as many octets as the cidr width specifies, and these\n> are the only required octets on input (though extra .0's can be specified).\n\nCool. You answered another question I had before I asked it.\n\n> > I think we are just about there. If we go with my plan (completely\n> > different functionality for now and fold it later) there should be\n> > no API change later. There will be code and catalogue changes but\n> > they should be relatively painless.\n> \n> so shall i test the inet_cidr_ functions and punt them on in?\n\nOk, before I have a reality shift, the inet_cidr_ functions are simply\nthe original inet_net_ functions renamed, right?\n\nI so hate it when the universe does that stuttering thing.\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Mon, 19 Oct 1998 15:56:10 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Taral\n> Can't we just use a CONSTRAINT where a host address is expected? That sounds\n> easier than setting up two different types to me...\n\nThe constraint would be pretty complicated and it doesn't handle the\ndifferent output rules.\n\nDon't worry. After things settle down we'll fold things together so\nthat there is two input wrapper functions and everything else will be\nhandled by the same functions so you won't hardly know the difference.\nI too originally thought there should be one type but Paul has convinced\nme otherwise.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Mon, 19 Oct 1998 16:00:46 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> bigness of names doesn't matter. applications matter. i can see a use for\n> both types, but they are inherently different types. a host that has a\n> netmask which can be expressed in cidr notation is one such type. a net\n> that has a netmask which must be expressed in cidr notation is another such\n> type. the difference comes down to \"host part must be zero\" for the network\n> type. there are also some minor differences in the input/output formats,\n> since a host address always has four octets on both input and output, while\n> a network only prints as many octets as the cidr width specifies, and these\n> are the only required octets on input (though extra .0's can be specified).\n> \n> > I think we are just about there. If we go with my plan (completely\n> > different functionality for now and fold it later) there should be\n> > no API change later. There will be code and catalogue changes but\n> > they should be relatively painless.\n> \n> so shall i test the inet_cidr_ functions and punt them on in?\n\nYep. Those are good points. Let's go.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 19:59:17 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> And besides:\n> \n> if (strlen(\"D'Arcy J.M. Cain\") > strlen(\"Paul Vixie\"))\n> printf(\"Who's the big name now? :-)\\n\");\n> \n> And look at all that StUdLy punctuation. :-)\n\nThat was funny.\n\n> > > I think we are just about there. If we go with my plan (completely\n> > > different functionality for now and fold it later) there should be\n> > > no API change later. There will be code and catalogue changes but\n> > > they should be relatively painless.\n> > \n> > so shall i test the inet_cidr_ functions and punt them on in?\n> \n> Ok, before I have a reality shift, the inet_cidr_ functions are simply\n> the original inet_net_ functions renamed, right?\n> \n> I so hate it when the universe does that stuttering thing.\n\nJust a clarification. If you have function that can be called with\neither type, you can put two entries in pg_proc(or I can :-) ), and that\nway the one function can take two types as parameters. Allows use to\nmaintain the type easier, and less bloat.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 20:03:28 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> > > so shall i test the inet_cidr_ functions and punt them on in?\n> > \n> > Ok, before I have a reality shift, the inet_cidr_ functions are simply\n> > the original inet_net_ functions renamed, right?\n> > \n> > I so hate it when the universe does that stuttering thing.\n> \n> Just a clarification. If you have function that can be called with\n> either type, you can put two entries in pg_proc(or I can :-) ), and that\n> way the one function can take two types as parameters. Allows use to\n> maintain the type easier, and less bloat.\n\nMy idea is to add a flag to the inet structure that specifies whether\nthe object is a cidr or an inet. Then every function checks that to\ndecide what action to perform (which might be to raise an exception,\ntrying to apply the host function to a cidr type for example) with\none exception, the input function. I see this as the functions to\ndo input.\n\ninet *\ninet_common_in(char *src, int flag);\nThis takes the input string and a flag that says whether a cidr or\nan inet is being input. It pallocs the structure, sets the flag and\nperforms the conversion based the type. returns the inet pointer.\n\ninet *\ninet_net_in(char *src)\n{\n return inet_common_in(src, 0);\n}\n\ninet *\ninet_cidr_in(char *src)\n{\n return inet_common_in(src, 1);\n}\n\nHowever, let's get the two types in right now with two separate groups\nof functions and fold them after the release. It won't change the\nuser interface. Unless we think we can do it quickly.\n\nIn any case, maybe we can add the flag now since we figure we'll need\nit later anyway.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Mon, 19 Oct 1998 22:17:27 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> Thus spake Taral\n> > Can't we just use a CONSTRAINT where a host address is expected? That sounds\n> > easier than setting up two different types to me...\n> \n> The constraint would be pretty complicated and it doesn't handle the\n> different output rules.\n> \n> Don't worry. After things settle down we'll fold things together so\n> that there is two input wrapper functions and everything else will be\n> handled by the same functions so you won't hardly know the difference.\n> I too originally thought there should be one type but Paul has convinced\n> me otherwise.\n> \n\nIf you define a fuction that can take inet or cidr type, I recommend you\ndefine a typedef called something like \"inet_or_cidr\" that is the same\nas inet, and use that in functions that can take either type.\n\nYou can then clearly see what functions can take either type. Also, you\nwill not know at the time you are called what type is really being\npassed, but it may not matter, or you may be able to figure out what to\ndo based on the data inside the type.\n\nAlso, D'Arcy, the pressure is on. There will not be a lot of time for\ndebugging and redesign. I am here if you need help.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 23:00:20 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> However, let's get the two types in right now with two separate groups\n> of functions and fold them after the release. It won't change the\n> user interface. Unless we think we can do it quickly.\n> \n> In any case, maybe we can add the flag now since we figure we'll need\n> it later anyway.\n\nOnce we put entries in the system tables, those really are not going to\nchange dramatically until 6.5.\n\nAlso, at this point, the least amount of code that can be added, should\nbe added. We already _had_(broken) an inet type, and now we are going\nto be throwing a lot of duplicate code in there for a new type.\n\nThere is already concern that we are too close to the 6.4 final date to\ndo anything with the INET type. I am hearing that from another\ndeveloper.\n\nI am not sure what to advise, but adding a new type is not trivial. It\nis going to require an initdb by everyone, because it is going to be in\nthe regression test.\n\nNew type is involved, especially if I have to add unique indexing\nfunctions and other stuff for the new type.\n\nIf people really want the INET/CIDR type for 6.4, we are going to need\ntremendous effort to pull this off. That means good, clean code,\ndocumenation, and testing, regression tests, and soon.\n\nWe can not just throw this in, and we can't expect the entire tree to\nwait for a new type.\n\nMy personal opinion is that I am not ready to add a new type, and new\nduplicate functions for that type, this close to final. I can add the\ntype, and the pg_proc/indexing pointers to link in the existing\ninet functions, but full type inclusion is too much, I think.\n\nFor example, I have an inet_ops entry in pg_class. I don't want to add\nan cidr_ops function that behaves exactly the same. If we can't do this\nright, then we will not do it for 6.4. My experience is that dumping\npartial solutions into 250k lines of code is a bad thing.\n\nSo, if people really want it, it has to be _good_. If is not that\nimportant, it can wait.\n\nThese are my opinions, and of course, can be over-ruled.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 23:56:47 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> If you define a fuction that can take inet or cidr type, I recommend you\n> define a typedef called something like \"inet_or_cidr\" that is the same\n> as inet, and use that in functions that can take either type.\n\nPerhaps. We can look at that 6.4++. For now there will be redundancy.\nAfterwards we'll look at how to fold them.\n\n> passed, but it may not matter, or you may be able to figure out what to\n> do based on the data inside the type.\n\nUltimately, I think the latter.\n\n> Also, D'Arcy, the pressure is on. There will not be a lot of time for\n> debugging and redesign. I am here if you need help.\n\nThe minute that Paul's code is in the tree, I'm there. However, if it\nhelps, here is cidr.c which is just the original inet.c with substitutions.\nThere may be minor changes when I see Paul's functions so I won't submit it\nto patches yet but if someone wants to review it early, here it is.\n\n\n/*\n *\tPostgreSQL type definitions for the CIDR type. This\n *\tis for IP V4 CIDR notation, but prepared for V6: just\n *\tadd the necessary bits where the comments indicate.\n *\n *\t$Id$\n */\n\n#include <sys/types.h>\n#include <sys/socket.h>\n\n#include <stdio.h>\n#include <string.h>\n#include <errno.h>\n\n#include <netinet/in.h>\n#include <arpa/inet.h>\n\n#include <postgres.h>\n#include <utils/palloc.h>\n#include <utils/builtins.h>\n#include <utils/inet.h>\n\nstatic int v4bitncmp(unsigned int a1, unsigned int a2, int bits);\n\n/*\n *\tAccess macros.\tAdd IPV6 support.\n */\n\n#define ip_addrsize(inetptr) \\\n\t(((inet_struct *)VARDATA(inetptr))->family == AF_INET ? 4 : -1)\n\n#define ip_family(inetptr) \\\n\t(((inet_struct *)VARDATA(inetptr))->family)\n\n#define ip_bits(inetptr) \\\n\t(((inet_struct *)VARDATA(inetptr))->bits)\n\n#define ip_v4addr(inetptr) \\\n\t(((inet_struct *)VARDATA(inetptr))->addr.ipv4_addr)\n\n/*\n *\tIP address reader.\n */\n\ninet *\ncidr_in(char *src)\n{\n\tint\t\t\tbits;\n\tinet\t *dst;\n\n\tdst = palloc(VARHDRSZ + sizeof(inet_struct));\n\tif (dst == NULL)\n\t{\n\t\telog(ERROR, \"unable to allocate memory in cidr_in()\");\n\t\treturn (NULL);\n\t}\n\t/* First, try for an IP V4 address: */\n\tip_family(dst) = AF_INET;\n\tbits = inet_cidr_pton(ip_family(dst), src, &ip_v4addr(dst), ip_addrsize(dst));\n\tif ((bits < 0) || (bits > 32))\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"could not parse \\\"%s\\\"\", src);\n\t\tpfree(dst);\n\t\treturn (NULL);\n\t}\n\tVARSIZE(dst) = VARHDRSZ\n\t\t+ ((char *) &ip_v4addr(dst) - (char *) VARDATA(dst))\n\t\t+ ip_addrsize(dst);\n\tip_bits(dst) = bits;\n\treturn (dst);\n}\n\n/*\n *\tIP address output function.\n */\n\nchar *\ncidr_out(inet *src)\n{\n\tchar\t *dst,\n\t\t\t\ttmp[sizeof(\"255.255.255.255/32\")];\n\n\tif (ip_family(src) == AF_INET)\n\t{\n\t\t/* It's an IP V4 address: */\n\t\tif (inet_cidr_ntop(AF_INET, &ip_v4addr(src), ip_addrsize(src), ip_bits(src),\n\t\t\t\t\t\t tmp, sizeof(tmp)) < 0)\n\t\t{\n\t\t\telog(ERROR, \"unable to print address (%s)\", strerror(errno));\n\t\t\treturn (NULL);\n\t\t}\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"unknown address family (%d)\", ip_family(src));\n\t\treturn (NULL);\n\t}\n\tdst = palloc(strlen(tmp) + 1);\n\tif (dst == NULL)\n\t{\n\t\telog(ERROR, \"unable to allocate memory in cidr_out()\");\n\t\treturn (NULL);\n\t}\n\tstrcpy(dst, tmp);\n\treturn (dst);\n}\n\n/*\n *\tBoolean tests for magnitude. Add V4/V6 testing!\n */\n\nbool\ncidr_lt(inet *a1, inet *a2)\n{\n\tif ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET))\n\t{\n\t\tint\t\t\torder = v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2));\n\n\t\treturn ((order < 0) || ((order == 0) && (ip_bits(a1) < ip_bits(a2))));\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"cannot compare address families %d and %d\",\n\t\t\t ip_family(a1), ip_family(a2));\n\t\treturn (FALSE);\n\t}\n}\n\nbool\ncidr_le(inet *a1, inet *a2)\n{\n\treturn (inet_lt(a1, a2) || inet_eq(a1, a2));\n}\n\nbool\ncidr_eq(inet *a1, inet *a2)\n{\n\tif ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET))\n\t{\n\t\treturn ((ip_bits(a1) == ip_bits(a2))\n\t\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a1)) == 0));\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"cannot compare address families %d and %d\",\n\t\t\t ip_family(a1), ip_family(a2));\n\t\treturn (FALSE);\n\t}\n}\n\nbool\ncidr_ge(inet *a1, inet *a2)\n{\n\treturn (inet_gt(a1, a2) || inet_eq(a1, a2));\n}\n\nbool\ncidr_gt(inet *a1, inet *a2)\n{\n\tif ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET))\n\t{\n\t\tint\t\t\torder = v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2));\n\n\t\treturn ((order > 0) || ((order == 0) && (ip_bits(a1) > ip_bits(a2))));\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"cannot compare address families %d and %d\",\n\t\t\t ip_family(a1), ip_family(a2));\n\t\treturn (FALSE);\n\t}\n}\n\nbool\ncidr_ne(inet *a1, inet *a2)\n{\n\treturn (!inet_eq(a1, a2));\n}\n\nbool\ncidr_sub(inet *a1, inet *a2)\n{\n\tif ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET))\n\t{\n\t\treturn ((ip_bits(a1) > ip_bits(a2))\n\t\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2)) == 0));\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"cannot compare address families %d and %d\",\n\t\t\t ip_family(a1), ip_family(a2));\n\t\treturn (FALSE);\n\t}\n}\n\nbool\ncidr_subeq(inet *a1, inet *a2)\n{\n\tif ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET))\n\t{\n\t\treturn ((ip_bits(a1) >= ip_bits(a2))\n\t\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2)) == 0));\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"cannot compare address families %d and %d\",\n\t\t\t ip_family(a1), ip_family(a2));\n\t\treturn (FALSE);\n\t}\n}\n\nbool\ncidr_sup(inet *a1, inet *a2)\n{\n\tif ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET))\n\t{\n\t\treturn ((ip_bits(a1) < ip_bits(a2))\n\t\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a1)) == 0));\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"cannot compare address families %d and %d\",\n\t\t\t ip_family(a1), ip_family(a2));\n\t\treturn (FALSE);\n\t}\n}\n\nbool\ncidr_supeq(inet *a1, inet *a2)\n{\n\tif ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET))\n\t{\n\t\treturn ((ip_bits(a1) <= ip_bits(a2))\n\t\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a1)) == 0));\n\t}\n\telse\n\t{\n\t\t/* Go for an IPV6 address here, before faulting out: */\n\t\telog(ERROR, \"cannot compare address families %d and %d\",\n\t\t\t ip_family(a1), ip_family(a2));\n\t\treturn (FALSE);\n\t}\n}\n\n/*\n *\tComparison function for sorting. Add V4/V6 testing!\n */\n\nint4\ncidr_cmp(inet *a1, inet *a2)\n{\n\tif (ntohl(ip_v4addr(a1)) < ntohl(ip_v4addr(a2)))\n\t\treturn (-1);\n\telse if (ntohl(ip_v4addr(a1)) > ntohl(ip_v4addr(a2)))\n\t\treturn (1);\n\treturn 0;\n}\n\n/*\n *\tBitwise comparison for V4 addresses. Add V6 implementation!\n */\n\nstatic int\nv4bitncmp(unsigned int a1, unsigned int a2, int bits)\n{\n\tunsigned long mask = 0;\n\tint\t\t\ti;\n\n\tfor (i = 0; i < bits; i++)\n\t\tmask = (mask >> 1) | 0x80000000;\n\ta1 = ntohl(a1);\n\ta2 = ntohl(a2);\n\tif ((a1 & mask) < (a2 & mask))\n\t\treturn (-1);\n\telse if ((a1 & mask) > (a2 & mask))\n\t\treturn (1);\n\treturn (0);\n}\n\n\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Tue, 20 Oct 1998 00:11:26 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > If you define a fuction that can take inet or cidr type, I recommend you\n> > define a typedef called something like \"inet_or_cidr\" that is the same\n> > as inet, and use that in functions that can take either type.\n> \n> Perhaps. We can look at that 6.4++. For now there will be redundancy.\n> Afterwards we'll look at how to fold them.\n\nI hope you realize that that 6.4++ would have to be 6.5. We typically\ndon't to system table changes as part of minor releases, unless there is\na really good reason. It is hard to do for us and for the users.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 00:21:10 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> > no. the last three inputs are not valid where a host address is expected.\n> \n> Can you amplify? Is it correct as far as cidr goes? If so, I have no\n> problem making it an error for the inet type. My thinking was based\n> on the earlier concept of having one type and accepting networks in it.\n> If we have the separate cidr type then I guess inet should always require\n> 4 octets (until ipv6 anyway) and cidr should be used for networks.\n\nNetworks do not have to have all four octets specified, only enough octets\nto cover the prefix length that's given. Networks should have default\nnetmasks based on classful assumptions. Networks never have any bits beyond\ntheir prefix length, which is why the question of \"nonzero host part\" does\nnot even really arise. There is no \"host part\" in a network. And in my\nconcept of them, they are variably sized.\n\n> How about something like 192.63.0.0/16? Should that be an error under the\n> inet type since it is the network? I am thinking not since technically\n> 192.63.0.0 is a valid host under 192.63/16 although it is generally\n> avoided since there is still software that assumes that it is the\n> network or even the broadcast.\n\nThe all-zeros host address is available and value, but as you say, deprecated.\n\n> > so shall i test the inet_cidr_ functions and punt them on in?\n> \n> Ok, before I have a reality shift, the inet_cidr_ functions are simply\n> the original inet_net_ functions renamed, right?\n\nNo. The inet_net_ functions describe CIDRized networks. The inet_cidr_\nfunctions, which I've written but not yet committed to BIND 8.next, are\nnew. I am absolutely loathe to submit them at this point, since the act\nof bridging host addresses and their prefix lengths makes no sense to me.\n\nHere's a cisco showing a CIDR block (mine, as it turns out):\n\n\tpalo-alto>sho ip rou 204.152.184.0\n\tRouting entry for 204.152.184.0/21, supernet\n\t Known via \"bgp 1280\", distance 20, metric 0\n\t Tag 3557, type external\n\t Last update from 198.32.176.3 2w0d ago\n\t Routing Descriptor Blocks:\n\t * 198.32.176.3, from 198.32.176.3, 2w0d ago\n\t Route metric is 0, traffic share count is 1\n\t AS Hops 1\n\nHere's a BSD/OS box showing a bunch of CIDR blocks (inside my network):\n\n\t# netstat -rn\n\tDestination Gateway Flags Refs Use Interface\n\tdefault 204.152.184.4 UG 0 103154609 de1\n\t127 127.0.0.1 UR 0 0 lo0\n\t127.0.0.1 127.0.0.1 UH 0 59294 lo0\n\t192.5.5.1 204.152.184.19 UGH 0 1160628 de0\n\t192.5.5.2 204.152.184.19 UGH 0 507879 de0\n\t192.5.5.88/29 204.152.184.19 UG 0 4 de0\n\t192.5.5.96/27 204.152.184.19 UG 0 35150 de0\n\t192.5.5.124/30 204.152.184.19 UG 0 12361 de0\n\t192.5.5.241 204.152.184.4 UGH 0 55164 de1\n\t198.32.176 204.152.184.1 UG 0 15250 de1\n\t198.32.176.6 204.152.184.1 UGHc 0 76 de1\n\t204.152.184/28 link#2 UC 0 0 de1\n\t204.152.184.1 0:c0:95:e0:1e:1c UHLc 4 493 de1\n\t204.152.184.3 0:c0:95:e0:2e:8c UHLc 0 1 lo0\n\t204.152.184.4 0:c0:95:e0:1e:24 UHLc 4 7125 de1\n\t204.152.184.5 0:c0:95:e0:26:80 UHLc 1 0 de1\n\t204.152.184.16/29 link#1 UC 0 0 de0\n\t^C\n\nThe things which are \"hosts\" have four octets, are of fixed length, and do not\nhave netmasks. The things which are \"networks\" have some other number of\noctets, are variably sized, and do have netmasks (actually, prefix lengths).\n\nAt the risk of reopening a discussion which ought to have been closed a long\nwhile back, I suggest that polymorphism is bad and that an INET type which is\nbuilt out of BIND's inet_net_ functions is perfectly capable of holding a\nhost address (since /32 is the default if all four octets are specified),\nand that we run absolutely zero risk if we call the type INET, add indexing\nlinkages for it, and get outta town.\n\nThere is an argument for a sugartype called IHOST which generates a /32 INET\nbut uses inet_pton() and inet_ntop() for parsing/printing so as to avoid the\n\"/%d\" that nobody needs to see and also to ensure that only fully formed host\naddresses can be input.\n\nThere is an argument for another, completely separate type, called MACADDR,\nwhich is an IEEE 48-bit address (suitable for ethernet or fddi arp tables).\n\nThere is no argument I know of for a type which combines the host address and\nthe netmask of its interface. I've got applications in mind for each of the\nabove three types -- has anybody got a specific application they want to build\nwhich requires a CIDR-like host address which is of fixed size, has the\nability to preserve the \"unset\"-ness of the prefix length from input through\nstorage to output, and also has the ability to preserve a particular prefix\nlength from input through storage to output?\n\nI'll use INET in a registry database like IANA's or InterNIC's.\n\nI'll use INET, IHOST and MACADDR in a distributed DHCP database.\n\nWhat would anybody use a mixture of INET and IHOST for, that they could not\ndo just as easily with a pair of IHOST's?\n\nForget the number theory for a moment and let's talk about applications which\nare uniquely enabled by any new type we consider. Once that's done, we can\ntalk about avoiding unfortunate overlaps.\n\nI've got the code done for supporting hosts-with-prefixes, but I don't like it\nand I would not use it in any PgSQL application I can imagine. Help?\n",
"msg_date": "Mon, 19 Oct 1998 23:21:37 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "> There is already concern that we are too close to the 6.4 final date to\n> do anything with the INET type. I am hearing that from another\n> developer.\n\nYes.\n\n> I am not sure what to advise, but adding a new type is not trivial. It\n> is going to require an initdb by everyone, because it is going to be in\n> the regression test.\n\nI propose that we rename CIDR to INET, base it on the existing inet_net_*\nfunctions, and have done with it. We can add IHOST next time.\n\n> My personal opinion is that I am not ready to add a new type, and new\n> duplicate functions for that type, this close to final. I can add the\n> type, and the pg_proc/indexing pointers to link in the existing\n> inet functions, but full type inclusion is too much, I think.\n\nI don't know how to help with this.\n\n> For example, I have an inet_ops entry in pg_class. I don't want to add\n> an cidr_ops function that behaves exactly the same. If we can't do this\n> right, then we will not do it for 6.4. My experience is that dumping\n> partial solutions into 250k lines of code is a bad thing.\n\nYes.\n\n> So, if people really want it, it has to be _good_. If is not that\n> important, it can wait.\n\nI believe that I am to blame for the last minute nature of this, because I\nwas not properly focused on applications during the much earlier discussion.\n\nBecause we're at the end of our time, I propose that we rename the type to\nINET, use the existing inet_net_ functions, and blow the bolts.\n",
"msg_date": "Mon, 19 Oct 1998 23:25:25 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "Thus spake Paul A Vixie\n> Networks do not have to have all four octets specified, only enough octets\n> to cover the prefix length that's given. Networks should have default\n> netmasks based on classful assumptions. Networks never have any bits beyond\n> their prefix length, which is why the question of \"nonzero host part\" does\n> not even really arise. There is no \"host part\" in a network. And in my\n> concept of them, they are variably sized.\n\nI think we (at least I) are still confused about what the types are. I\nsee two types on the table;\n\n a) A type to hold networks only\n b) A type to hold hosts with optional netmasks\n\nCan we agree that these are the two types we are discussing? If so can we\nagree on which is which? I have been assuming that A was cidr and B was\ninet but perhaps I have it backwards.\n\n> The all-zeros host address is available and value, but as you say, deprecated.\n\nBut not illegal, right?\n\n> No. The inet_net_ functions describe CIDRized networks. The inet_cidr_\n> functions, which I've written but not yet committed to BIND 8.next, are\n> new. I am absolutely loathe to submit them at this point, since the act\n> of bridging host addresses and their prefix lengths makes no sense to me.\n\nOK, so it looks like I had it backwards all along. So we can just put the\ninet.c function back in exactly as it was then? If we agree on this then\nwe are half way there and at least we have something for 6.4 right now.\n\n> At the risk of reopening a discussion which ought to have been closed a long\n> while back, I suggest that polymorphism is bad and that an INET type which is\n> built out of BIND's inet_net_ functions is perfectly capable of holding a\n> host address (since /32 is the default if all four octets are specified),\n> and that we run absolutely zero risk if we call the type INET, add indexing\n> linkages for it, and get outta town.\n> \n> There is an argument for a sugartype called IHOST which generates a /32 INET\n> but uses inet_pton() and inet_ntop() for parsing/printing so as to avoid the\n> \"/%d\" that nobody needs to see and also to ensure that only fully formed host\n> addresses can be input.\n\nSo host only - no additional information carried in the type?\n\n> There is an argument for another, completely separate type, called MACADDR,\n> which is an IEEE 48-bit address (suitable for ethernet or fddi arp tables).\n\nNothing to do with our current discussion, right? You are just mentioning\nit for completeness and we shall never speak of it again, or at least not\ntill after 6.4?\n\n> There is no argument I know of for a type which combines the host address and\n> the netmask of its interface. I've got applications in mind for each of the\n> above three types -- has anybody got a specific application they want to build\n> which requires a CIDR-like host address which is of fixed size, has the\n> ability to preserve the \"unset\"-ness of the prefix length from input through\n> storage to output, and also has the ability to preserve a particular prefix\n> length from input through storage to output?\n\nRadius server for Ascend terminal servers.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Tue, 20 Oct 1998 09:27:51 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> > However, let's get the two types in right now with two separate groups\n> > of functions and fold them after the release. It won't change the\n> > user interface. Unless we think we can do it quickly.\n> > \n> > In any case, maybe we can add the flag now since we figure we'll need\n> > it later anyway.\n> \n> Once we put entries in the system tables, those really are not going to\n> change dramatically until 6.5.\n\nWhile I would like everything in there now, I can live with this as long\nas the user interface doesn't change.\n\nHow about this? We now have an inet type pretty much completed. Let's\nput that back in right now. copy all the files involved, substitute\ninet to cidr in the function names and make it the new cidr type. Once\nwe agree on which one is the host+netmask type add the extra functions\nfor netmask, masklen, host, network_without_bits, network_with_bits and\nbroadcast making them stubs if necessary. I'll get my code into the\nright file as soon as possible. Later we can fold things in better\nbut we can do it without changing the catalogues. If it makes it more\nefficient we can change the catalogue where 6.5 comes out.\n\nIf we can do this right away I think we have a good chance of getting\nit into 6.4 properly anyway.\n\n> My personal opinion is that I am not ready to add a new type, and new\n> duplicate functions for that type, this close to final. I can add the\n> type, and the pg_proc/indexing pointers to link in the existing\n> inet functions, but full type inclusion is too much, I think.\n> \n> For example, I have an inet_ops entry in pg_class. I don't want to add\n> an cidr_ops function that behaves exactly the same. If we can't do this\n> right, then we will not do it for 6.4. My experience is that dumping\n> partial solutions into 250k lines of code is a bad thing.\n\nOf course this is your decision. We have the inet type now and as long\nas we know what that type is, we can always add the other for 6.5 if\nwe can't get it in now.\n\n> So, if people really want it, it has to be _good_. If is not that\n> important, it can wait.\n\nSo far less than a half dozen people are really involved in this discussion.\nHow about a quick straw poll of who really wants this in?\n\n> These are my opinions, and of course, can be over-ruled.\n\nNot if you need to do the work on the catalogues they can't. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n\n",
"msg_date": "Tue, 20 Oct 1998 09:51:21 -0400 (EDT)",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> If people really want the INET/CIDR type for 6.4, we are going to need\n> tremendous effort to pull this off. That means good, clean code,\n> documenation, and testing, regression tests, and soon.\n\nWould it be helpful to discuss this in a more direct matter? I am sitting\nin #postgreSQL on EFNet (IRC) if anyone wants to join me there.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Tue, 20 Oct 1998 10:11:32 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> I think we (at least I) are still confused about what the types are. I\n> see two types on the table;\n> \n> a) A type to hold networks only\n> b) A type to hold hosts with optional netmasks\n> \n> Can we agree that these are the two types we are discussing? If so can we\n> agree on which is which? I have been assuming that A was cidr and B was\n> inet but perhaps I have it backwards.\n\nI am challenging the (b) type above. I see three types:\n\n\tINET\tholds CIDRized networks\n\tIHOST\tholds host end-addresses\n\tMACADDR\tholes IEEE 48-bit ether/fddi addresses\n\nAs to whether the type names are backwards, probably not. It's only if we\npostulate type (b) above that an ambiguity arises as to whether the type we\nwere calling CIDR refers to hosts+netmasks or to CIDRized networks.\n\nCertainly the new inet_cidr_ functions were written for the former while the\nexisting inet_net_ functions were written for the latter.\n\n> > The all-zeros host address is available ..., but as you say, deprecated.\n> \n> But not illegal, right?\n\nNope, just unusable on Suns.\n\n> > No. The inet_net_ functions describe CIDRized networks. The inet_cidr_\n> > functions, which I've written but not yet committed to BIND 8.next, are\n> > new. I am absolutely loathe to submit them at this point, since the act\n> > of bridging host addresses and their prefix lengths makes no sense to me.\n> \n> OK, so it looks like I had it backwards all along. So we can just put the\n> inet.c function back in exactly as it was then? If we agree on this then\n> we are half way there and at least we have something for 6.4 right now.\n\nAgreed.\n\n> > There is an argument for a sugartype called IHOST which generates a /32\n> > INET but uses inet_pton() and inet_ntop() for parsing/printing so as to\n> > avoid the \"/%d\" that nobody needs to see and also to ensure that only\n> > fully formed host addresses can be input.\n> \n> So host only - no additional information carried in the type?\n\nThat would be my preference. But as it would be the same underlying type,\nit would be possible to ask for all supernet INETs of some IHOST -- the\nsupernet/subnet comparison functions would be inherently polymorphic. I've\nalready got an application in mind that would benefit from this polymorphism.\n\n> > There is an argument for another, completely separate type, called MACADDR,\n> > which is an IEEE 48-bit address (suitable for ethernet or fddi arp tables).\n> \n> Nothing to do with our current discussion, right? You are just mentioning\n> it for completeness and we shall never speak of it again, or at least not\n> till after 6.4?\n\nI'd thought that the fellow who wrote ip_and_mac had already submitted the\nMACADDR type. If not, then clearly it is way too late to consider it for 6.4.\n\n> > ... -- has anybody got a specific application they want to build which\n> > requires a CIDR-like host address which is of fixed size, has the ability\n> > to preserve the \"unset\"-ness of the prefix length from input through\n> > storage to output, and also has the ability to preserve a particular prefix\n> > length from input through storage to output?\n> \n> Radius server for Ascend terminal servers.\n\nIs there no way to accomplish this without efficiency loss using a pair of\nIHOSTs, one for the host address and one for the netmask?\n\nI note that polymorphism of the supernet/subnet tests between (a) and (b) as\ndenoted in the top of the text I've quoted in this message is a lot more work\nthan the inherent polymorphism of supernet/subnet tests between an INET and\nan IHOST as I've proposed them here.\n",
"msg_date": "Tue, 20 Oct 1998 08:51:05 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "\nI'm voting that the rest of the peanut gallary sit down and allow\nPostgreSQL to adopt Vixie's world vision of the INET type. If after the\nrelease it is found to be lacking it can be addressed then.\n\nLets give the man some credit; he's got a better idea of what will be\nuseful than anyone else I've seen here. Everyone else seems to be\nstumbling around attempting to argue a point while trying to figure out\nwhat the heck they're arguing about. If you don't understand what Paul is\ntrying to do, then sit down, read a few books and come back when you do.\n\nThis whole issue is really heading towards a stalemate otherwise.\n\nAs usual, feel free to ignore me.\n\nPaul, thanks for taking the time to educate everyone here. Your patience\nis appriciated and I only hope that the features you need will pan out in\nthe release.\n\nOn Mon, 19 Oct 1998, Paul A Vixie wrote:\n\n> > > no. the last three inputs are not valid where a host address is expected.\n> > \n> > Can you amplify? Is it correct as far as cidr goes? If so, I have no\n> > problem making it an error for the inet type. My thinking was based\n> > on the earlier concept of having one type and accepting networks in it.\n> > If we have the separate cidr type then I guess inet should always require\n> > 4 octets (until ipv6 anyway) and cidr should be used for networks.\n> \n> Networks do not have to have all four octets specified, only enough octets\n> to cover the prefix length that's given. Networks should have default\n> netmasks based on classful assumptions. Networks never have any bits beyond\n> their prefix length, which is why the question of \"nonzero host part\" does\n> not even really arise. There is no \"host part\" in a network. And in my\n> concept of them, they are variably sized.\n> \n> > How about something like 192.63.0.0/16? Should that be an error under the\n> > inet type since it is the network? I am thinking not since technically\n> > 192.63.0.0 is a valid host under 192.63/16 although it is generally\n> > avoided since there is still software that assumes that it is the\n> > network or even the broadcast.\n> \n> The all-zeros host address is available and value, but as you say, deprecated.\n> \n> > > so shall i test the inet_cidr_ functions and punt them on in?\n> > \n> > Ok, before I have a reality shift, the inet_cidr_ functions are simply\n> > the original inet_net_ functions renamed, right?\n> \n> No. The inet_net_ functions describe CIDRized networks. The inet_cidr_\n> functions, which I've written but not yet committed to BIND 8.next, are\n> new. I am absolutely loathe to submit them at this point, since the act\n> of bridging host addresses and their prefix lengths makes no sense to me.\n> \n> Here's a cisco showing a CIDR block (mine, as it turns out):\n> \n> \tpalo-alto>sho ip rou 204.152.184.0\n> \tRouting entry for 204.152.184.0/21, supernet\n> \t Known via \"bgp 1280\", distance 20, metric 0\n> \t Tag 3557, type external\n> \t Last update from 198.32.176.3 2w0d ago\n> \t Routing Descriptor Blocks:\n> \t * 198.32.176.3, from 198.32.176.3, 2w0d ago\n> \t Route metric is 0, traffic share count is 1\n> \t AS Hops 1\n> \n> Here's a BSD/OS box showing a bunch of CIDR blocks (inside my network):\n> \n> \t# netstat -rn\n> \tDestination Gateway Flags Refs Use Interface\n> \tdefault 204.152.184.4 UG 0 103154609 de1\n> \t127 127.0.0.1 UR 0 0 lo0\n> \t127.0.0.1 127.0.0.1 UH 0 59294 lo0\n> \t192.5.5.1 204.152.184.19 UGH 0 1160628 de0\n> \t192.5.5.2 204.152.184.19 UGH 0 507879 de0\n> \t192.5.5.88/29 204.152.184.19 UG 0 4 de0\n> \t192.5.5.96/27 204.152.184.19 UG 0 35150 de0\n> \t192.5.5.124/30 204.152.184.19 UG 0 12361 de0\n> \t192.5.5.241 204.152.184.4 UGH 0 55164 de1\n> \t198.32.176 204.152.184.1 UG 0 15250 de1\n> \t198.32.176.6 204.152.184.1 UGHc 0 76 de1\n> \t204.152.184/28 link#2 UC 0 0 de1\n> \t204.152.184.1 0:c0:95:e0:1e:1c UHLc 4 493 de1\n> \t204.152.184.3 0:c0:95:e0:2e:8c UHLc 0 1 lo0\n> \t204.152.184.4 0:c0:95:e0:1e:24 UHLc 4 7125 de1\n> \t204.152.184.5 0:c0:95:e0:26:80 UHLc 1 0 de1\n> \t204.152.184.16/29 link#1 UC 0 0 de0\n> \t^C\n> \n> The things which are \"hosts\" have four octets, are of fixed length, and do not\n> have netmasks. The things which are \"networks\" have some other number of\n> octets, are variably sized, and do have netmasks (actually, prefix lengths).\n> \n> At the risk of reopening a discussion which ought to have been closed a long\n> while back, I suggest that polymorphism is bad and that an INET type which is\n> built out of BIND's inet_net_ functions is perfectly capable of holding a\n> host address (since /32 is the default if all four octets are specified),\n> and that we run absolutely zero risk if we call the type INET, add indexing\n> linkages for it, and get outta town.\n> \n> There is an argument for a sugartype called IHOST which generates a /32 INET\n> but uses inet_pton() and inet_ntop() for parsing/printing so as to avoid the\n> \"/%d\" that nobody needs to see and also to ensure that only fully formed host\n> addresses can be input.\n> \n> There is an argument for another, completely separate type, called MACADDR,\n> which is an IEEE 48-bit address (suitable for ethernet or fddi arp tables).\n> \n> There is no argument I know of for a type which combines the host address and\n> the netmask of its interface. I've got applications in mind for each of the\n> above three types -- has anybody got a specific application they want to build\n> which requires a CIDR-like host address which is of fixed size, has the\n> ability to preserve the \"unset\"-ness of the prefix length from input through\n> storage to output, and also has the ability to preserve a particular prefix\n> length from input through storage to output?\n> \n> I'll use INET in a registry database like IANA's or InterNIC's.\n> \n> I'll use INET, IHOST and MACADDR in a distributed DHCP database.\n> \n> What would anybody use a mixture of INET and IHOST for, that they could not\n> do just as easily with a pair of IHOST's?\n> \n> Forget the number theory for a moment and let's talk about applications which\n> are uniquely enabled by any new type we consider. Once that's done, we can\n> talk about avoiding unfortunate overlaps.\n> \n> I've got the code done for supporting hosts-with-prefixes, but I don't like it\n> and I would not use it in any PgSQL application I can imagine. Help?\n> \n\n-- \n| Matthew N. Dodd | 78 280Z | 75 164E | 84 245DL | FreeBSD/NetBSD/Sprite/VMS |\n| [email protected] | This Space For Rent | ix86,sparc,m68k,pmax,vax |\n| http://www.jurai.net/~winter | Are you k-rad elite enough for my webpage? |\n\n",
"msg_date": "Tue, 20 Oct 1998 12:10:08 -0400 (EDT)",
"msg_from": "\"Matthew N. Dodd\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "> > There is already concern that we are too close to the 6.4 final date to\n> > do anything with the INET type. I am hearing that from another\n> > developer.\n> \n> Yes.\n> \n> > I am not sure what to advise, but adding a new type is not trivial. It\n> > is going to require an initdb by everyone, because it is going to be in\n> > the regression test.\n> \n> I propose that we rename CIDR to INET, base it on the existing inet_net_*\n> functions, and have done with it. We can add IHOST next time.\n\nNo need. It is already called INET.\n\n> \n> > My personal opinion is that I am not ready to add a new type, and new\n> > duplicate functions for that type, this close to final. I can add the\n> > type, and the pg_proc/indexing pointers to link in the existing\n> > inet functions, but full type inclusion is too much, I think.\n> \n> I don't know how to help with this.\n\nI will do the work, but am hesitant to do too much system table fiddling\nat this point.\n\n> \n> > For example, I have an inet_ops entry in pg_class. I don't want to add\n> > an cidr_ops function that behaves exactly the same. If we can't do this\n> > right, then we will not do it for 6.4. My experience is that dumping\n> > partial solutions into 250k lines of code is a bad thing.\n> \n> Yes.\n> \n> > So, if people really want it, it has to be _good_. If is not that\n> > important, it can wait.\n> \n> I believe that I am to blame for the last minute nature of this, because I\n> was not properly focused on applications during the much earlier discussion.\n> \n> Because we're at the end of our time, I propose that we rename the type to\n> INET, use the existing inet_net_ functions, and blow the bolts.\n\nFortunately, it is already called INET. We have all the system catalogs\nwired for the type. We just need working versions of some of those\nfunctions, and D'Arcy can go at it.\n\nIf we are good, this may not even require an initdb, because the changes\nwere made long ago for the INET type.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 12:40:35 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> > > There is an argument for another, completely separate type, called MACADDR,\n> > > which is an IEEE 48-bit address (suitable for ethernet or fddi arp tables).\n> > \n> > Nothing to do with our current discussion, right? You are just mentioning\n> > it for completeness and we shall never speak of it again, or at least not\n> > till after 6.4?\n> \n> I'd thought that the fellow who wrote ip_and_mac had already submitted the\n> MACADDR type. If not, then clearly it is way too late to consider it for 6.4.\n> \n\n>From psql's \\dT command:\n\n\tmacaddr |MAC address \n\tinet |IP address \n\n> > > ... -- has anybody got a specific application they want to build which\n> > > requires a CIDR-like host address which is of fixed size, has the ability\n> > > to preserve the \"unset\"-ness of the prefix length from input through\n> > > storage to output, and also has the ability to preserve a particular prefix\n> > > length from input through storage to output?\n> > \n> > Radius server for Ascend terminal servers.\n> \n> Is there no way to accomplish this without efficiency loss using a pair of\n> IHOSTs, one for the host address and one for the netmask?\n> \n> I note that polymorphism of the supernet/subnet tests between (a) and (b) as\n> denoted in the top of the text I've quoted in this message is a lot more work\n> than the inherent polymorphism of supernet/subnet tests between an INET and\n> an IHOST as I've proposed them here.\n\nI can't comment on the rest of this, because I don't understand it.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 12:53:35 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Paul A Vixie\n> > I think we (at least I) are still confused about what the types are. I\n> > see two types on the table;\n> > \n> > a) A type to hold networks only\n> > b) A type to hold hosts with optional netmasks\n> > \n> > Can we agree that these are the two types we are discussing? If so can we\n> > agree on which is which? I have been assuming that A was cidr and B was\n> > inet but perhaps I have it backwards.\n> \n> I am challenging the (b) type above. I see three types:\n> \n> \tINET\tholds CIDRized networks\n\nSo that's the type a above.\n\n> \tIHOST\tholds host end-addresses\n> \tMACADDR\tholes IEEE 48-bit ether/fddi addresses\n> \n> As to whether the type names are backwards, probably not. It's only if we\n> postulate type (b) above that an ambiguity arises as to whether the type we\n> were calling CIDR refers to hosts+netmasks or to CIDRized networks.\n\nOriginally I thought we were calling 'a' the cidr type and 'b' the inet\ntype hence my confusion. I still think that that is the better but since\nwe have working code and it is already named, I guess we should go with it.\n\nFor convenience, let's call my 'b' type above a CHOST type for CIDRized\nhost. So hopefully we all have the types straight.\n\n INET holds CIDRized networks\n IHOST holds host end-addresses\n CHOST host plus netmask.\n\n> Certainly the new inet_cidr_ functions were written for the former while the\n> existing inet_net_ functions were written for the latter.\n\nSo it is designed for CHOST then? Well, you know my opinion.\n\n> > > The all-zeros host address is available ..., but as you say, deprecated.\n> > \n> > But not illegal, right?\n> \n> Nope, just unusable on Suns.\n\n<TANGENT>Right. I wonder if we should consider making this a compile-time\noption or something post-6.4.</TANGENT>\n\n> > > No. The inet_net_ functions describe CIDRized networks. The inet_cidr_\n> > > functions, which I've written but not yet committed to BIND 8.next, are\n> > > new. I am absolutely loathe to submit them at this point, since the act\n> > > of bridging host addresses and their prefix lengths makes no sense to me.\n> > \n> > OK, so it looks like I had it backwards all along. So we can just put the\n> > inet.c function back in exactly as it was then? If we agree on this then\n> > we are half way there and at least we have something for 6.4 right now.\n> \n> Agreed.\n\nSo unless someone is violently opposed then Bruce, can you put it back to\nthe way it was before I submitted my patch to inet.c? That gives us the\nINET type as defined above.\n\n> > So host only - no additional information carried in the type?\n> \n> That would be my preference. But as it would be the same underlying type,\n> it would be possible to ask for all supernet INETs of some IHOST -- the\n> supernet/subnet comparison functions would be inherently polymorphic. I've\n> already got an application in mind that would benefit from this polymorphism.\n\nYou think it should be a differnt type then? You can do it with one if\nyou use /32 for hosts, right? In fact, make a naked ip imply /32 for\nINET type but /-1 for CHOST type (if we go with it.)\n\nIn fact, forget the -1 idea. Default both types to /32 and never print\nthe bits for the CHOST type. That simplifies the calculations.\n\n> I'd thought that the fellow who wrote ip_and_mac had already submitted the\n> MACADDR type. If not, then clearly it is way too late to consider it for 6.4.\n\nActually, I don't even know if that is in or not. In any case, it either\nis or isn't. We don't need to dwell on it. (So why am I? :-) )\n\n> > > ... -- has anybody got a specific application they want to build which\n> > > requires a CIDR-like host address which is of fixed size, has the ability\n> > > to preserve the \"unset\"-ness of the prefix length from input through\n> > > storage to output, and also has the ability to preserve a particular prefix\n> > > length from input through storage to output?\n> > \n> > Radius server for Ascend terminal servers.\n> \n> Is there no way to accomplish this without efficiency loss using a pair of\n> IHOSTs, one for the host address and one for the netmask?\n\nIt becomes messy. In fact, I would use an integer for the netmask in that\nsituation.\n\n> I note that polymorphism of the supernet/subnet tests between (a) and (b) as\n> denoted in the top of the text I've quoted in this message is a lot more work\n> than the inherent polymorphism of supernet/subnet tests between an INET and\n> an IHOST as I've proposed them here.\n\nThat's why I'm saying get rid of the -1 hack that I originally proposed.\n\nSo, INET is in. The question is what to do with IHOST and CHOST. I\nthink CHOST should go in. You can use it for IHOST too.\n\nI'm going to talk to Bruce in IRC now. I have another idea for making\none type for all 3 but I'm not sure of some details of the type system.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Tue, 20 Oct 1998 13:01:45 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "Thus spake Matthew N. Dodd\n> I'm voting that the rest of the peanut gallary sit down and allow\n> PostgreSQL to adopt Vixie's world vision of the INET type. If after the\n> release it is found to be lacking it can be addressed then.\n\nActually, we already have. The discussion is only over adding a second\ntype and what, exactly, it should be.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Tue, 20 Oct 1998 13:43:30 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
},
{
"msg_contents": "> Paul, thanks for taking the time to educate everyone here. Your patience\n> is appriciated and I only hope that the features you need will pan out in\n> the release.\n\nBe careful here. My purported credentials in the networking field do not\nnecessarily translate to expertise in the database field. I am a student\nof computer languages, especially their type systems. INET can be seen as\na supertype of (which means it can contain all the values) of IHOST. But\nthere is no subtype or supertype relationship between CHOST and either INET\nor IHOST. Since Postgres is a very typey database, I am attracted to the\nsimple approach of creating INET (which is done now?), later on creating\nIHOST as a subtype of it with different parsing/printing functions, and\nusing a pair of IHOSTs for any application which needs what CHOST does.\n\nThat's the computer language typing system approach, and the networking\napproach. I'm confident in those fields. I am less so in the database\nfield -- the last database application I was paid to write was in 1981,\nand that database was not even relational let alone object oriented.\n\nIHOST and INET is how I'd implement this functionality in, say, Modula-3.\nIt also makes for a relatively compact and unambiguous (that is, easy to\ngenerate a complete set of unit tests for) implementation. But it's very\npossible that I'm just an uneducated dolt when it comes to relational\nobject oriented databases and that anyone with more experience than I have\nwould automatically gravitate toward something like CHOST.\n\nQuoting Mike Smith, \"I am only an egg.\"\n",
"msg_date": "Tue, 20 Oct 1998 10:43:56 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "> Originally I thought we were calling 'a' the cidr type and 'b' the inet\n> type hence my confusion. I still think that that is the better but since\n> we have working code and it is already named, I guess we should go with it.\n\nThis sounds like consensus to me. Bruce said the same.\n\n> > > So host only - no additional information carried in the type?\n> > \n> > That would be my preference. But as it would be the same underlying type,\n> > it would be possible to ask for all supernet INETs of some IHOST -- the\n> > supernet/subnet comparison functions would be inherently polymorphic. I've\n> > already got an application in mind that would benefit from this\n> > polymorphism.\n> \n> You think it should be a differnt type then? You can do it with one if\n> you use /32 for hosts, right? In fact, make a naked ip imply /32 for\n> INET type but /-1 for CHOST type (if we go with it.)\n\nIn IHOST as I proposed it, it would have the same on-disk format as INET,\nbut with a fixed /32 and with different parsing and printing functions:\nthe parser would object unless all four octets and no /## was specified,\nand the printer would just print the octets and elide the /32. There are\nfunctions in BIND, i.e., inet_pton() and inet_ntop(), which do that kind\nof parsing and printing.\n\n> In fact, forget the -1 idea. Default both types to /32 and never print\n> the bits for the CHOST type. That simplifies the calculations.\n\nIf we're never printing the bits for CHOST, it's not different from IHOST?\n\n> > Is there no way to accomplish this without efficiency loss using a pair of\n> > IHOSTs, one for the host address and one for the netmask?\n> \n> It becomes messy. In fact, I would use an integer for the netmask in that\n> situation.\n\n\"Messy\" is not as strong a concern as performance, though, is it? If the\nonly time we need a host address and a netmask together is when a Radius\nserver using an SQL backend had to do some string arithmetic before sending\nthe Radius reply, then that's not as compelling an argument as it might be.\n",
"msg_date": "Tue, 20 Oct 1998 10:51:25 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind "
},
{
"msg_contents": "Thus spake Paul A Vixie\n> > You think it should be a differnt type then? You can do it with one if\n> > you use /32 for hosts, right? In fact, make a naked ip imply /32 for\n> > INET type but /-1 for CHOST type (if we go with it.)\n> \n> In IHOST as I proposed it, it would have the same on-disk format as INET,\n> but with a fixed /32 and with different parsing and printing functions:\n> the parser would object unless all four octets and no /## was specified,\n> and the printer would just print the octets and elide the /32. There are\n> functions in BIND, i.e., inet_pton() and inet_ntop(), which do that kind\n> of parsing and printing.\n> \n> > In fact, forget the -1 idea. Default both types to /32 and never print\n> > the bits for the CHOST type. That simplifies the calculations.\n> \n> If we're never printing the bits for CHOST, it's not different from IHOST?\n\nI meant default to /32 if it isn't specified. If it is then use what is\nspecified and print it if it isn't 32. So:\n\nInput Stored as Output as\n========= ========= =========\n198.96.119.226/28 198.96.119.226/28 198.96.119.226/28\n198.96.119.226/32 198.96.119.226/32 198.96.119.226\n198.96.119.226 198.96.119.226/32 198.96.119.226\n\nThe middle one is the hairy one. Using -1 for unspecified let's us\noutput /32 if that was what was input. If we can live with not\nprinting /32 ever then we can simplify the handling.\n\n> > > Is there no way to accomplish this without efficiency loss using a pair of\n> > > IHOSTs, one for the host address and one for the netmask?\n> > \n> > It becomes messy. In fact, I would use an integer for the netmask in that\n> > situation.\n> \n> \"Messy\" is not as strong a concern as performance, though, is it? If the\n> only time we need a host address and a netmask together is when a Radius\n> server using an SQL backend had to do some string arithmetic before sending\n> the Radius reply, then that's not as compelling an argument as it might be.\n\nActually, what is stored in the database (using CHOST) is exactly what\nthe terminal server needs - no calculations. To be complete I would\nneed to use the following to store everything I need without CHOST.\n\nnetwork::inet\ngateway::ihost\nmasklen::integer\n\nNow I guess I could get the masklen from the network but even then it\nisn't straightforward since I would have to do string operations on\nit to extract the bits after the slash. Either way it now becomes\npossible for the two to get out of sync so now I have to also check\nthe network and gateway to make sure that the latter is within the\nformer every time I use it.\n\nWith CHOST I just select the field and pass it to the Radius function\nthat called me.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Tue, 20 Oct 1998 16:29:17 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
}
] |
[
{
"msg_contents": "It's been quite some time since it was necessary to run the postmaster\nwith a specific TZ setting in order to do the regression tests. I see\nthat the instructions to change TZ have been removed from install.sgml\n(though not yet from the text INSTALL file). But the instructions still\ntell you to start a postmaster for the regression test, then kill it and\nstart another one for production use. Seems to me we could just\nrearrange the order of the steps: start the postmaster normally, *then*\nrun the regression tests if you feel like it. Is there a reason not to\nsimplify the instructions that way?\n\n\t\t\tregards, tom lane\n\nPS: Actually, what we really need is an easy way to run the regression\ntests on a new build without having to disturb the production\ninstallation until you know the new build works. (Everyone else does\n\"make test\" *before* \"make install\" ... with pgsql you have to do it the\nother way around. Not good for mission-critical software.)\n\nRight now I think this requires configuring, building, installing with\nnonstandard values of POSTGRESDIR and PGPORT, then running the\nregression test in that environment, then (if successful) throwing away\nall that work and repeating the build with standard configuration so you\ncan install it live. Bleah. And you still run the risk of blowing it\nby misconfiguring the second time around.\n\nWhat I really want to be able to run the regression test on the software\nsitting in the build tree, without doing an install at all. Any\nthoughts on what it would take to do that?\n",
"msg_date": "Sun, 18 Oct 1998 13:35:29 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Anyone object to simplifying INSTALL instructions?"
}
] |
[
{
"msg_contents": "I just finished a major bout of makefile hacking. We have seven or\nso different makefiles that try to produce shared libraries, and only\ntwo of them (libpq and libpgtcl) have been really properly maintained\nover the last couple of months. I fixed the recently introduced \noverwrite-library-with-symlink bug in those two, and then tried to\npropagate what they know about shlibs into the other makefiles.\n\nWhat I have now builds and installs, so I went ahead and checked it in\n(what we had yesterday did not install correctly, so this can't be any\nworse...). But I'm thinking that this is still pretty messy and\nunmaintainable. In particular there is way too much replication of\nstuff between the different makefiles.\n\nI am thinking that we could improve matters by moving the common\nmaterial into a new file, say src/Makefile.shlib, that would be\nincluded by each existing makefile that wants to build a shared lib.\nThe common file could contain all of the platform-specific variable\nsettings as well as some of the common rules (AFAIC the rules for\nbuilding and installing the shlib could be in the common file...).\n\nDoes anyone have an objection to this?\n\n\nWhile we're on the subject of build procedures: a lot of the makefiles\ncontain cruft for building ordinary unshared libraries like this:\n\nlibpq.a: $(OBJS)\nifdef MK_NO_LORDER\n\t$(AR) $(AROPT) libpq.a $(OBJS) \nelse\n\t$(AR) $(AROPT) libpq.a `lorder $(OBJS) | tsort`\nendif\n\t$(RANLIB) libpq.a\n\nI don't believe there are any Unix platforms left that still need the\nlorder|tsort dance --- that's been superseded by ranlib on every\nplatform that has a linker that needs sorted modules within a library.\n(libjpeg, which I believe to be considerably more widely-spread than\nPostgres, has never done anything but a plain ranlib call --- and I\nhave never yet heard a complaint.)\n\nThe lorder stuff is evidently giving someone a hard time, since the\nMK_NO_LORDER cruft has recently been introduced to allow it to be turned\noff.\n\nI propose just ripping out the lorder|tsort calls. Can anyone point\nto a platform on which they are still needed?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sun, 18 Oct 1998 15:59:56 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "shared library makefile changes checked in"
}
] |
[
{
"msg_contents": "Hi,\n\n PL/pgSQL is still enabled in the default build. I think it\n would be good to have the same semantics for\n\n gram.c\n y.tab.h\n scan.c\n\n as we have them in backend/parser (don't remove them on make\n clean and ship ready to compile versions in the distribution\n that are older than gram.y and scan.l). All that must be done\n is to remove them from the rm line in the clean step in\n plpgsql/src/Makefile.in and add the files to CVS.\n\n And I don't know if the same is/isn't already arranged for\n ecpg.\n\n BTW: an sgml documentation for PL/pgSQL and PL/Tcl will soon\n be available.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Sun, 18 Oct 1998 22:00:45 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "PL/pgSQL lex/yacc produced files"
}
] |
[
{
"msg_contents": "Outch,\n\njust got the following on the 6.4b2:\n\npgsql=> create table t1 (a int4, b date);\nCREATE\npgsql=> insert into t1 values (1, 'now');\nINSERT 18282 1\npgsql=> update t1 set b = NULL where a = 1;\npqReadData() -- backend closed the channel unexpectedly.\n This probably means the backend terminated abnormally before or while processing the request.\nWe have lost the connection to the backend, so further processing is impossible. Terminating.\n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n",
"msg_date": "Sun, 18 Oct 1998 23:13:49 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Crashing backend!"
},
{
"msg_contents": "[email protected] (Jan Wieck) writes:\n> just got the following on the 6.4b2:\n> pgsql=> create table t1 (a int4, b date);\n> CREATE\n> pgsql=> insert into t1 values (1, 'now');\n> INSERT 18282 1\n> pgsql=> update t1 set b = NULL where a = 1;\n> pqReadData() -- backend closed the channel unexpectedly.\n\nFWIW, that sequence works fine for me here, using current CVS sources.\nPerhaps it's platform-dependent? (I'm on HPUX.) Or maybe a bug fixed\nsince the 6.4b2 tarball?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sun, 18 Oct 1998 19:39:22 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Crashing backend! "
},
{
"msg_contents": "> Outch,\n> \n> just got the following on the 6.4b2:\n> \n> pgsql=> create table t1 (a int4, b date);\n> CREATE\n> pgsql=> insert into t1 values (1, 'now');\n> INSERT 18282 1\n> pgsql=> update t1 set b = NULL where a = 1;\n> pqReadData() -- backend closed the channel unexpectedly.\n> This probably means the backend terminated abnormally before or while processing the request.\n> We have lost the connection to the backend, so further processing is impossible. Terminating.\n\nJust tested it on BSDI. Worked fine.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 01:49:12 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Crashing backend!"
},
{
"msg_contents": "Jan Wieck wrote:\n> Outch,\n> \n> just got the following on the 6.4b2:\n> \n> pgsql=> create table t1 (a int4, b date);\n> CREATE\n> pgsql=> insert into t1 values (1, 'now');\n> INSERT 18282 1\n> pgsql=> update t1 set b = NULL where a = 1;\n> pqReadData() -- backend closed the channel unexpectedly.\n> This probably means the backend terminated abnormally before or while processing the request.\n> We have lost the connection to the backend, so further processing is impossible. Terminating.\n\nI just tested it on the UnixWare 7.x port and it worked fine.\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Mon, 19 Oct 1998 02:16:56 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Crashing backend! "
},
{
"msg_contents": ">\n> Outch,\n>\n> just got the following on the 6.4b2:\n>\n> pgsql=> create table t1 (a int4, b date);\n> CREATE\n> pgsql=> insert into t1 values (1, 'now');\n> INSERT 18282 1\n> pgsql=> update t1 set b = NULL where a = 1;\n> pqReadData() -- backend closed the channel unexpectedly.\n> This probably means the backend terminated abnormally before or while processing the request.\n> We have lost the connection to the backend, so further processing is impossible. Terminating.\n\n Seems something is spying my email here on the system. As\n soon as you all reported that it works for you the problem\n disappeared.\n\n Sorry for the panic.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 19 Oct 1998 12:07:38 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Crashing backend!"
}
] |
[
{
"msg_contents": "\nTwo seperate questions..\n\n1) Is there any reason why I can't create a unique index on an int8?\n\n It tells me this: \n ERROR: Can't find a default operator class for type 20.\n\n\n2) Are there any more changes expected in the code for libpq?\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Sun, 18 Oct 1998 17:42:34 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": true,
"msg_subject": "unique indexes and pqlib"
},
{
"msg_contents": "Vince Vielhaber <[email protected]> writes:\n> 1) Is there any reason why I can't create a unique index on an int8?\n> It tells me this: \n> ERROR: Can't find a default operator class for type 20.\n\nint8 is new and not very thoroughly supported yet. Probably there\njust isn't support for making an index on it (Thomas?).\n\n> 2) Are there any more changes expected in the code for libpq?\n\nEver? Yes. Before 6.4 release? Perhaps not, but I wouldn't swear to\nit. We are at a bug-fixes-only stage of the release cycle, but I do not\npromise there are no bugs in libpq.\n\nWhy do you care, exactly? You shouldn't need to worry whether a\nlibrary you use is changing internally, as long as the API is stable\n(and I certainly hope we do not find bugs that force an API change\nat this point...)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sun, 18 Oct 1998 19:11:11 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] unique indexes and pqlib "
},
{
"msg_contents": "\nOn 18-Oct-98 Tom Lane wrote:\n> Vince Vielhaber <[email protected]> writes:\n>> 2) Are there any more changes expected in the code for libpq?\n> \n> Ever? Yes. Before 6.4 release? Perhaps not, but I wouldn't swear to\n> it. We are at a bug-fixes-only stage of the release cycle, but I do not\n> promise there are no bugs in libpq.\n> \n> Why do you care, exactly? You shouldn't need to worry whether a\n> library you use is changing internally, as long as the API is stable\n> (and I certainly hope we do not find bugs that force an API change\n> at this point...)\n\nI'm doing the libpq and psql ports for OS/2. \n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Sun, 18 Oct 1998 19:23:49 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] unique indexes and pqlib"
},
{
"msg_contents": "> > 1) Is there any reason why I can't create a unique index on an int8?\n> > It tells me this:\n> > ERROR: Can't find a default operator class for type 20.\n> int8 is new and not very thoroughly supported yet. Probably there\n> just isn't support for making an index on it (Thomas?).\n\nI'll try to look at it. But I'm starting to get swamped with last-minute\nbig additions to the documentation, and will have only limited time for\nother stuff.\n\nbtw, I haven't looked yet but I hope the odbc makefile stuff still can\nbuild standalone. Hopefully we are converging on an acceptable set of\nMakefiles; I'd made changes recently - which you'd noticed broke\ninstallations which did not do major.minor version numbering on shared\nlibraries - in response to major breakage introduced for some other\nreason which damaged installations _with_ major.minor version numbering.\nThough despite appearances my response wasn't intended to rebreak the\nother systems :/\n\n - Tom\n",
"msg_date": "Mon, 19 Oct 1998 07:07:45 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] unique indexes and pqlib"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> btw, I haven't looked yet but I hope the odbc makefile stuff still can\n> build standalone.\n\nI didn't intentionally break that --- in fact I added some stuff to\nodbc's private configure script to ensure it would still support the\nMakefile. But I don't know enough about ODBC to test it, so I'd prefer\nto let someone who knows what it's for make sure that it still builds\nand works. (I can only say that it does now build and install a valid\nlibrary on HPUX, which it didn't before.)\n\n> Hopefully we are converging on an acceptable set of\n> Makefiles; I'd made changes recently - which you'd noticed broke\n> installations which did not do major.minor version numbering on shared\n> libraries - in response to major breakage introduced for some other\n> reason which damaged installations _with_ major.minor version\n> numbering.\n\nRight. Actually the recent changes were very good --- they made it\nfeasible to have a large chunk of the sharedlib build rules be textually\nidentical in different makefiles. I just took the next logical step\nand moved the textually identical stuff into a single file. Should be\na lot easier to maintain this way.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 19 Oct 1998 10:05:42 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Recent Makefile changes (was Re: unique indexes and pqlib)"
},
{
"msg_contents": "> \n> TwoBB seperate questions..\n> \n> 1) Is there any reason why I can't create a unique index on an int8?\n> \n> It tells me this: \n> ERROR: Can't find a default operator class for type 20.\n\nThis will be fixed in 6.5, due to go into beta soon.\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 15 Mar 1999 09:12:17 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] unique indexes and pqlib"
}
] |
[
{
"msg_contents": "\nOn 19-Oct-98 Tom Lane wrote:\n> Vince Vielhaber <[email protected]> writes:\n>>> Why do you care, exactly?\n> \n>> I'm doing the libpq and psql ports for OS/2. \n> \n> Ah. In that case, pleased to meetcha. I guess you can either wait till\n> 6.4 formal release (maybe another couple weeks?), or get involved now.\n> An advantage of getting involved now is that any porting patches you find\n> necessary could still be fed back into the 6.4 release. Do you care how\n> much your version diverges from 6.4 official?\n\nI'll be getting involved now (even tho I've been here for a few months),\nthat should make things alot easier.\n\n> I've somehow unofficially acquired the position of head libpq basher,\n> so feel free to talk to me about any changes you need there.\n> \n> regards, tom lane\n\nGreat! Around the time of 6.2 (I think), Marc submitted a couple of changes\nfor me, mostly minor. Unless major changes have occurred I'm hoping it'll\ngo smoothly and rather uneventful as 6.3 did. \n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Sun, 18 Oct 1998 20:32:19 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] unique indexes and pqlib"
},
{
"msg_contents": "Vince Vielhaber <[email protected]> writes:\n> Great! Around the time of 6.2 (I think), Marc submitted a couple of changes\n> for me, mostly minor. Unless major changes have occurred I'm hoping it'll\n> go smoothly and rather uneventful as 6.3 did. \n\nWell, actually, I've done some pretty major damage to the internals of\nlibpq since 6.3.2 ;-). The Windows guys report that porting it to\nWindows wasn't hard, but unless OS/2 offers something very close to\nBerkeley sockets and the select() system call, you might be in for\ntrouble. Let me know how it goes.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 19 Oct 1998 09:47:53 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] unique indexes and pqlib "
}
] |
[
{
"msg_contents": "> I want to confirm about OFFSET.\n> (Sorry. There are so many mails for me to check that I can't follow \n> what(where) it is now.)\n> \n> I think OFFSET was the offset for indices when \n> \n> \tSELECT *\n> \tFROM tab\n> \tWHERE col1 < some_unknown_value\n> \tORDER BY col1\n> \tUSING INDEX tab_idx(452) COUNT 100\n> or\n> \tSET INDEX TO tab_idx START 452 COUNT 100\n> \n> by Bruce\n> \n> The definition of OFFSET was changed to mean the offset for resultset\n> (the set of target rows(without OFFSET)) now ?\n> \n> such as \n> SELECT ... [LIMIT count [, offset]]\n> \n> by Jan\n> \n> Thanks.\n\n\nYes, old meaning was index size to scan, new meaning is result size to\nscan.\n\nThe old and new meaning are the same if there is an index matching the\nORDER BY. The old meaning was not workable if there was not matching\nindex, which is bad. In a join situation, the old meaning offered a few\nmore optimization possibilities, because you could do the OFFSET on the\nindex rather than computing the join, then the OFFSET. However, the\nnumber of result rows could vary in a join situation, that that is weird\nfor people. Add to that there is no logical way to know you have\nreached the end, because a list of 100 index entries could return zero\nrows. That is also bad. And I don't think people are going to be\nseeking into the middle of large result sets very often, so the new\nmeaning of counting the result rows is the best way to go.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 01:54:51 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] What about LIMIT in SELECT ?"
}
] |
[
{
"msg_contents": "Hi,\n\nI just tried to compile Sundays snapshot. I get this error:\n\nistrat.c 496: [error]: 2324 Undefined: 'F_OIDEQ'\n\nI greped through the includes, however I couldn't find a definition\nfor it. Did I download a \"work in progress\" snapshot?\n\nOr might it already be fixed in Monday's snapshot?\n--\ncu\n\tFrank\n",
"msg_date": "Mon, 19 Oct 1998 10:38:52 +0200 (MDT)",
"msg_from": "Frank Ridderbusch <[email protected]>",
"msg_from_op": true,
"msg_subject": "Missing define F_OIDEQ in Sundays snapshot"
}
] |
[
{
"msg_contents": "> -----Original Message-----\n> From: Jan Wieck [mailto:[email protected]]\n> Sent: Friday, October 16, 1998 5:21 PM\n> To: Hiroshi Inoue\n> Cc: [email protected]; [email protected]\n> Subject: Re: [HACKERS] What about LIMIT in SELECT ?\n> \n> \n> Hiroshi Inoue wrote:\n> \n> > In the following cases I didn't modify my code to use index scan,\n> > because I couldn't formulate how to tell PostgreSQL optimizer whether\n> > the response to get first rows is needed or the throughput to process\n> > sufficiently many target rows is needed.\n> >\n> > 3.The access plan made by current PostgreSQL optimizer for a query with\n> > ORDER BY clause doesn't include index scan.\n> >\n> > I thought the use of Tatsuo's QUERY_LIMIT to decide that the responce\n> > is needed. It is sufficient but not necessary ?\n> > In Oracle the hints FIRST_ROWS,ALL_ROWS are used.\n> \n> I still think that the QUERY LIMIT should be part of the\n> parse tree and not thrown in by a magic SET command. If\n> rewriting or function calls turn the one query sent to the\n> backend into multiple queries processed internal, how should\n> this QUERY LIMIT variable know to which of all the queries it\n> has to be applied? It can really break functions and rewrite\n> rules if this variable is used on all queries while it is\n> set.\n> \n> For your case 3 I think, if there is a QUERY LIMIT in the\n> parse tree, the (future) optimizer definitely knows that not\n> all rows will get processed even if there is no qualification\n> given. So if there is an index, that matches the ORDER BY\n> clause and it is no a join and the (future) executor handles\n> OFFSET in single table index scans fast, it could choose an\n> index scan for this query too.\n>\n\nWhen using cursors,in most cases the response to get first(next) rows \nis necessary for me,not the throughput.\nHow can we tell PostgreSQL optimzer that the response is necessary ?\n\nHiroshi Inoue\[email protected] \n\n",
"msg_date": "Mon, 19 Oct 1998 18:52:46 +0900",
"msg_from": "\"Hiroshi Inoue\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] What about LIMIT in SELECT ?"
},
{
"msg_contents": "Hiroshi Inoue wrote:\n\n> When using cursors,in most cases the response to get first(next) rows\n> is necessary for me,not the throughput.\n> How can we tell PostgreSQL optimzer that the response is necessary ?\n\n With my LIMIT patch, the offset and the row count are part of\n the querytree. And if a LIMIT is given, the limitCount elemet\n of the querytree (a Node *) isn't NULL what it is by default.\n\n When a LIMIT is given, the optimizer could assume that first\n rows is wanted (even if the limit is ALL maybe - but I have\n to think about this some more). And this assumption might let\n it decide to use an index to resolve an ORDER BY even if no\n qualification was given.\n\n Telling the optimizer that first rows wanted in a cursor\n operation would read\n\n DECLARE CURSOR c FOR SELECT * FROM mytab ORDER BY a LIMIT ALL;\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 19 Oct 1998 12:30:52 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] What about LIMIT in SELECT ?"
}
] |
[
{
"msg_contents": "\n> Bruce Momjian <[email protected]> writes:\n> > What if someone wants the rows from 500 to the end. Should we allow\n> > the syntax to be:\n> > SELECT ... [LIMIT count] [OFFSET offset]\n> > LIMIT and OFFSET are independent.\n> \n> I like that syntax the best, but remember we are not inventing in\n> a green field here. Isn't this a feature that already exists in\n> other DBMs? We should probably copy their syntax, unless it's\n> truly spectacularly awful...\n\nInformix uses the following syntax, but does not have the offset feature:\n\n\tselect first 100 * from table ....\n\nAndreas\n\t\n\n\n\n",
"msg_date": "Mon, 19 Oct 1998 12:19:56 +0200",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] SELECT ... LIMIT (trial implementation) "
},
{
"msg_contents": "> \n> > Bruce Momjian <[email protected]> writes:\n> > > What if someone wants the rows from 500 to the end. Should we allow\n> > > the syntax to be:\n> > > SELECT ... [LIMIT count] [OFFSET offset]\n> > > LIMIT and OFFSET are independent.\n> > \n> > I like that syntax the best, but remember we are not inventing in\n> > a green field here. Isn't this a feature that already exists in\n> > other DBMs? We should probably copy their syntax, unless it's\n> > truly spectacularly awful...\n> \n> Informix uses the following syntax, but does not have the offset feature:\n> \n> \tselect first 100 * from table ....\n\nI definitely don't like that syntax.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 11:27:30 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] SELECT ... LIMIT (trial implementation)"
}
] |
[
{
"msg_contents": "============================================================================\n POSTGRESQL BUG REPORT TEMPLATE\n============================================================================\n\nYour name: Jose' Soares\nYour email address:[email protected]\n\n\nSystem Configuration\n---------------------\n Architecture (example: Intel Pentium): Intel Pentium\n\n Operating System (example: Linux 2.0.26 ELF) : Linux 2.0.34 Elf\n\n PostgreSQL version (example: PostgreSQL-6.1) : PostgreSQL-6.4-BETA2\n\n Compiler used (example: gcc 2.7.2) : gcc 2.7.2.1\n\n\nPlease enter a FULL description of your problem:\n\nPostgreSQL tells me, the following query is not valid, but I can't see\nnothing wrong.\n\nprova=> select * from sp;\nsno |pno |qty\n-----+-----+---\nS1 |P1 |300\nS1 |P2 |200\nS1 |P3 |400\nS2 |P2 |400\nS3 |P2 |200\nS4 |P4 |300\nS4 |P5 |400\n(7 rows)\n\nprova=> select sno,qty from sp group by sno,qty having qty=300;\nERROR: This is not a valid having query!\n\nI tried the same query on Informix-SE and I have the following:\n\nsno qty\n\nS1 300.000\nS4 300.000\n\nI tried it also on Solid and I have the following:\n\nSOLID SQL Editor (teletype) v.02.20.0007\nselect sno,qty from sp group by sno,qty having qty = 300;\nSNO QTY\n--- ---\nS1 300.\nS4 300.\n2 rows fetched.\n\n Jose'\n\n\n",
"msg_date": "Mon, 19 Oct 1998 12:57:54 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": true,
"msg_subject": "HAVING bug"
},
{
"msg_contents": "> \n> PostgreSQL tells me, the following query is not valid, but I can't see\n> nothing wrong.\n> \n> prova=> select * from sp;\n> sno |pno |qty\n> -----+-----+---\n> S1 |P1 |300\n> S1 |P2 |200\n> S1 |P3 |400\n> S2 |P2 |400\n> S3 |P2 |200\n> S4 |P4 |300\n> S4 |P5 |400\n> (7 rows)\n> \n> prova=> select sno,qty from sp group by sno,qty having qty=300;\n> ERROR: This is not a valid having query!\n\nHAVING is only to be used with aggregates. In your case, qty is not an\naggregate.\n\nAt least, that is what I think is supposed to happen.\n\n> \n> I tried the same query on Informix-SE and I have the following:\n> \n> sno qty\n> \n> S1 300.000\n> S4 300.000\n> \n> I tried it also on Solid and I have the following:\n> \n> SOLID SQL Editor (teletype) v.02.20.0007\n> select sno,qty from sp group by sno,qty having qty = 300;\n> SNO QTY\n> --- ---\n> S1 300.\n> S4 300.\n> 2 rows fetched.\n> \n> Jose'\n> \n> \n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 11:29:56 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] HAVING bug"
},
{
"msg_contents": "> prova=> select sno,qty from sp group by sno,qty having qty=300;\n> ERROR: This is not a valid having query!\n\nUse WHERE for conditions like this.\n\nSELECT sno, qty FROM sp WHERE qty=300 GROUP BY sno;\n\nTaral\n",
"msg_date": "Mon, 19 Oct 1998 12:10:44 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] HAVING bug"
}
] |
[
{
"msg_contents": "Hello all,\n\nI have a table SP with a primary key SNO,\nI'm trying if v6.4beta2 uses indexes\nbut seems it uses indexes only with '=' operator:\n\nprova=> explain select * from sp where sno = 'S1';\nNOTICE: QUERY PLAN:\nIndex Scan using sp_pkey on sp (cost=0.00 size=0 width=28)\nEXPLAIN\n\nprova=> explain select * from sp where sno = 'S1' or sno = 'S2';\nNOTICE: QUERY PLAN:\nSeq Scan on sp (cost=0.00 size=0 width=28)\nEXPLAIN\n\nprova=> explain select * from sp where sno <> 'S1';\nNOTICE: QUERY PLAN:\nSeq Scan on sp (cost=0.00 size=0 width=28)\nEXPLAIN\n\nprova=> explain select * from sp where sno > 'S1';\nNOTICE: QUERY PLAN:\nSeq Scan on sp (cost=0.00 size=0 width=28)\nEXPLAIN\nprova=> explain select * from sp where sno < 'S1';\nNOTICE: QUERY PLAN:\nSeq Scan on sp (cost=0.00 size=0 width=28)\nEXPLAIN\n\n Jose'\n\n\n",
"msg_date": "Mon, 19 Oct 1998 13:09:58 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": true,
"msg_subject": "using indexes"
},
{
"msg_contents": "Sferacarta Software wrote:\n> \n> Hello all,\n> \n> I have a table SP with a primary key SNO,\n> I'm trying if v6.4beta2 uses indexes\n> but seems it uses indexes only with '=' operator:\n\nregression=> explain select unique1 from onek where unique1 < 3;\nNOTICE: QUERY PLAN:\n\nIndex Scan using onek_unique1 on onek (cost=28.67 size=334 width=4)\n\n> prova=> explain select * from sp where sno = 'S1';\n> NOTICE: QUERY PLAN:\n> Index Scan using sp_pkey on sp (cost=0.00 size=0 width=28)\n> EXPLAIN\n\nLooks like maybe you need to vacuum to get statistics for the optimizer?\n\n - Tom\n",
"msg_date": "Mon, 19 Oct 1998 14:46:59 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] using indexes"
},
{
"msg_contents": "Hi Tom,\n\n\nTGL> Jose' wrote:\n>> \n>> Hello all,\n>> \n>> I have a table SP with a primary key SNO,\n>> I'm trying if v6.4beta2 uses indexes\n>> but seems it uses indexes only with '=' operator:\n\nregression=>> explain select unique1 from onek where unique1 < 3;\nTGL> NOTICE: QUERY PLAN:\n\nTGL> Index Scan using onek_unique1 on onek (cost=28.67 size=334 width=4)\n\n>> prova=> explain select * from sp where sno = 'S1';\n>> NOTICE: QUERY PLAN:\n>> Index Scan using sp_pkey on sp (cost=0.00 size=0 width=28)\n>> EXPLAIN\n\nTGL> Looks like maybe you need to vacuum to get statistics for the optimizer?\n\nTGL> - Tom\n\n\nAfter vacuum the behavior is the same:\n\nprova=> create table onek(unique1 int4 primary key);\nNOTICE: CREATE TABLE/PRIMARY KEY will create implicit index onek_pkey\nfor table onek\nCREATE\nprova=> insert into onek values (1);\nINSERT 152858 1\nprova=> insert into onek values (10);\nINSERT 152859 1\nprova=> explain select unique1 from onek where unique1 < 3;\nNOTICE: QUERY PLAN:\n\nSeq Scan on onek (cost=0.00 size=0 width=4)\n\nEXPLAIN\nprova=> vacuum;\nVACUUM\nprova=> explain select unique1 from onek where unique1 < 3;\nNOTICE: QUERY PLAN:\n\nSeq Scan on onek (cost=1.07 size=1 width=4)\n\nEXPLAIN\n\nPS:\n What about:\n select unique1 from onek where unique1 = 3 OR unique1 = 1;\n ^^\n does it works on v6.4 ?\nJose'\n\n\n",
"msg_date": "Mon, 19 Oct 1998 18:12:49 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re[2]: [HACKERS] using indexes"
},
{
"msg_contents": "> PS:\n> What about:\n> select unique1 from onek where unique1 = 3 OR unique1 = 1;\n> ^^\n> does it works on v6.4 ?\n\nIt should.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 13:07:13 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Re[2]: [HACKERS] using indexes"
},
{
"msg_contents": "Bruce Momjian wrote:\n> \n> > PS:\n> > What about:\n> > select unique1 from onek where unique1 = 3 OR unique1 = 1;\n> > ^^\n> > does it works on v6.4 ?\n> \n> It should.\n\nThis is a great new.\nThis thing is absolutely neccessary to have better performances using\nODBC and ACCESS.\n\nThank you\nJose'\n",
"msg_date": "Tue, 20 Oct 1998 11:47:50 +0200",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] using indexes"
},
{
"msg_contents": "On Mon, 19 Oct 1998, Sferacarta Software wrote:\n\n> Hi Tom,\n> \n> \n> TGL> Jose' wrote:\n> >> \n> >> Hello all,\n> >> \n> >> I have a table SP with a primary key SNO,\n> >> I'm trying if v6.4beta2 uses indexes\n> >> but seems it uses indexes only with '=' operator:\n> \n> regression=>> explain select unique1 from onek where unique1 < 3;\n> TGL> NOTICE: QUERY PLAN:\n> \n> TGL> Index Scan using onek_unique1 on onek (cost=28.67 size=334 width=4)\n> \n> >> prova=> explain select * from sp where sno = 'S1';\n> >> NOTICE: QUERY PLAN:\n> >> Index Scan using sp_pkey on sp (cost=0.00 size=0 width=28)\n> >> EXPLAIN\n> \n> TGL> Looks like maybe you need to vacuum to get statistics for the optimizer?\n> \n> TGL> - Tom\n> \n> \n> After vacuum the behavior is the same:\n> \n> prova=> create table onek(unique1 int4 primary key);\n> NOTICE: CREATE TABLE/PRIMARY KEY will create implicit index onek_pkey\n> for table onek\n> CREATE\n> prova=> insert into onek values (1);\n> INSERT 152858 1\n> prova=> insert into onek values (10);\n> INSERT 152859 1\n> prova=> explain select unique1 from onek where unique1 < 3;\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on onek (cost=0.00 size=0 width=4)\n\n\tThe table has two records in it...why would it use an index?\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Tue, 20 Oct 1998 07:26:09 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Re[2]: [HACKERS] using indexes"
},
{
"msg_contents": "> Bruce Momjian wrote:\n> > \n> > > PS:\n> > > What about:\n> > > select unique1 from onek where unique1 = 3 OR unique1 = 1;\n> > > ^^\n> > > does it works on v6.4 ?\n> > \n> > It should.\n> \n> This is a great new.\n> This thing is absolutely neccessary to have better performances using\n> ODBC and ACCESS.\n\nYes. We worked hard on this because it was a much needed feature.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 12:46:34 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] using indexes"
}
] |
[
{
"msg_contents": "Hello!!!\n\nThis is a test. Sorry.\n\n-- \nThanx for paying your attention.\n\nAlexey.\nE-Mail: [email protected]\n*---={ A tout le monde }=---*\n",
"msg_date": "Mon, 19 Oct 1998 14:13:32 +0300 (EET DST)",
"msg_from": "Alexey Ugnichev <[email protected]>",
"msg_from_op": true,
"msg_subject": "Very sorry - testing..."
}
] |
[
{
"msg_contents": "Hi all,\n\nDoes anyone know if there are plans to upgrade the \nCOPY command so it spits back some intelligent results\nwhen it finishes. At the present time if it finds a bad\nrow or a non-unique key set, it simply quits out with a\nstatement:\n\n\t\"resetting connection\"\n\nWhich is not very informative. Let me know if this part of\nthe code needs working on. If nobody else is adding this,\nI would be happy to dive in.\n\nI could imagine this spitting out a message about a certain\nline not passing muster, or even the total number of rows\nin the file input/output being listed (as is now returned\nwith the INSERT/DELETE/UPDATE/SELECT commands).\n\n\nCheers,\n\nPaul\n-- \n Paul Friberg - [email protected] - http://www.isti.com\n (914) 256-9290 Office - (914) 256-9299 FAX\n ISTI==Instrumental Software Technologies, Inc.\nSystems Integration and Software Development Specialists\n",
"msg_date": "Mon, 19 Oct 1998 09:03:39 -0400",
"msg_from": "Paul Friberg CEO - ISTI <[email protected]>",
"msg_from_op": true,
"msg_subject": "PostgreSQL COPY command"
}
] |
[
{
"msg_contents": "Hi all,\n\nI created a table and a view on my database:\n\nCREATE TABLE prova (bp bpchar);\nCREATE VIEW wprova AS SELECT * FROM prova;\n\n\nIf I unload it using pg_dump I have this:\n\nCREATE TABLE \"prova\" (\"bp\" char(-5));\nCREATE TABLE \"wprova\" (\"bp\" char(-5));\nCOPY \"prova\" FROM stdin;\n\\.\ngetIndices(): SELECT failed\n\n\nNB: There are two errors:\n 1. My view \"wprova\" becomes a table.\n 2. The length of \"bp\" is wrong.\n\nI'm using v6.4beta2 on Linux.\n\nJose'\n\n\n",
"msg_date": "Mon, 19 Oct 1998 15:21:35 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": true,
"msg_subject": "pg_dump bug"
}
] |
[
{
"msg_contents": "> I tried to make sense of your message about the need for \"gcc -E -\"\n> in the autoconfigure results, but I couldn't get it to work.\n\nv6.4 will do the right thing if cpp is missing, and will use\n gcc -E -\ninstead.\n\nDo a \"find\" through the Makefiles and replace \"cpp\" with \"gcc -E -\" in\nplaces where it is executed as a command:\n\n cd $(POSTGRESDIR) # the top level directory\n find . -name '*akefile' -exec grep cpp {} \\; -print\n\nthen edit the files which are mentioned.\n\nbtw, I'm not sure why some versions of gcc are missing this. My rpm of\n2.7.2.1 had it, but 2.7.2.3 did not. egcs seems to be missing it also,\nso it looks like it may be a deprecated feature?\n\n - Tom\n",
"msg_date": "Mon, 19 Oct 1998 14:52:51 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: cpp preprocessor problem in postgres"
}
] |
[
{
"msg_contents": "FYI...\n\n> At http://nwalsh.com/docbook/dsssl/\n> \n> Version 1.19 includes support for several new languages, reworked\n> language support, and a number of enhancements/bug fixes. See\n> the ChangeLogs (http://nwalsh.com/docbook/dsssl/doc/) for more\n> info.\n\n - Tom\n",
"msg_date": "Mon, 19 Oct 1998 15:16:10 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "[Fwd: DocBook stylesheets 1.19 released]"
}
] |
[
{
"msg_contents": "Hi,\n\nI already asked in this mailing list but no comments yet :-)\nSomething wrong with cvs or it's my local problem ?\n\nmira:~/cvs$ cvs -z9 update pgsql\nFatal error, aborting.\n: no such user\nmira:~/cvs$ \n\n\n\tOleg\n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n",
"msg_date": "Mon, 19 Oct 1998 21:00:44 +0400 (MSD)",
"msg_from": "Oleg Bartunov <[email protected]>",
"msg_from_op": true,
"msg_subject": "cvs problem ?"
},
{
"msg_contents": "The same error message in Germany and no comments.\n\n-Egon\n\nOn Mon, 19 Oct 1998, Oleg Bartunov wrote:\n\n> Hi,\n> \n> I already asked in this mailing list but no comments yet :-)\n> Something wrong with cvs or it's my local problem ?\n> \n> mira:~/cvs$ cvs -z9 update pgsql\n> Fatal error, aborting.\n> : no such user\n> mira:~/cvs$ \n> \n> \n> \tOleg\n> \n> _____________________________________________________________\n> Oleg Bartunov, sci.researcher, hostmaster of AstroNet,\n> Sternberg Astronomical Institute, Moscow University (Russia)\n> Internet: [email protected], http://www.sai.msu.su/~megera/\n> phone: +007(095)939-16-83, +007(095)939-23-83\n> \n> \n> \n\n",
"msg_date": "Mon, 19 Oct 1998 19:12:46 +0200 (MET DST)",
"msg_from": "Egon Schmid <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] cvs problem ?"
}
] |
[
{
"msg_contents": "\n> \n> Hi,\n> \n> I already asked in this mailing list but no comments yet :-)\n> Something wrong with cvs or it's my local problem ?\n> \n> mira:~/cvs$ cvs -z9 update pgsql\n> Fatal error, aborting.\n> : no such user\n> mira:~/cvs$ \n\nI get the same thing here but look in the truss output.\n\nIsn't Marc getting a bit personal??\n\nKeith.\n\n\n a n o n c v s @ p o s t g r e s q l . o r g : / u s r / l o c a l\n / c v s r o o t A : 0 Z , I d Z\\n\nsend(4, 0x0005B0B8, 19, 0) = 19\n B E G I N A U T H R E Q U E S T\\n\nsend(4, 0x000822B0, 18, 0) = 18\n / u s r / l o c a l / c v s r o o t\nsend(4, \"\\n\", 1, 0) = 1\nsend(4, \" a n o n c v s\", 7, 0) = 7\nsend(4, \"\\n\", 1, 0) = 1\nsend(4, \" A : 0 Z , I d Z\", 9, 0) = 9\nsend(4, \"\\n\", 1, 0) = 1\nsend(4, 0x0005B0D0, 17, 0) = 17\n E N D A U T H R E Q U E S T\\n\nrecv(4, \" I\", 1, 0) = 1\nrecv(4, \" \", 1, 0) = 1\nrecv(4, \" L\", 1, 0) = 1\nrecv(4, \" O\", 1, 0) = 1\nrecv(4, \" V\", 1, 0) = 1\nrecv(4, \" E\", 1, 0) = 1\nrecv(4, \" \", 1, 0) = 1\nrecv(4, \" Y\", 1, 0) = 1\nrecv(4, \" O\", 1, 0) = 1\nrecv(4, \" U\", 1, 0) = 1\nrecv(4, \"\\n\", 1, 0) = 1\nfcntl(4, F_SETFD, 0x00000001) = 0\nfcntl(4, F_SETFD, 0x00000001) = 0\nfcntl(4, F_SETFD, 0x00000001) = 0\n \n> \tOleg\n> \n> _____________________________________________________________\n> Oleg Bartunov, sci.researcher, hostmaster of AstroNet,\n> Sternberg Astronomical Institute, Moscow University (Russia)\n> Internet: [email protected], http://www.sai.msu.su/~megera/\n> phone: +007(095)939-16-83, +007(095)939-23-83\n> \n> \n\n",
"msg_date": "Mon, 19 Oct 1998 18:28:15 +0100 (BST)",
"msg_from": "Keith Parks <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] cvs problem ?"
}
] |
[
{
"msg_contents": "I was hoping someone could shed some light on the \nfollowing problem:\n\nThe subselect under 6.4 ignores indexes built on \ntables. For example, consider the following two\ntables:\n\n\\d sales;\n\nTable = sales\n\nsupplysource varchar() not null 16\nsupply varchar() not null 16\nsupplyunit varchar() not null 2\nquantity float8 not null 8\ntarget varchar() not null 16\ncostcntr varchar() not null 8\nsaletype varchar() not null 16\nsaledate datetime not null 8\n\nIndices: k_sales\n k_sales_saledate\n k_sales_supply\n k_sales_target \n\n\n\\d locations;\n\nTable = locations\nlocation varchar() not null 16\ncostcntr varchar() not null 5\nsupplypath varchar() not null 64\nformpath varchar() not null 64\nengineerpath varchar() not null 64\n\nIndices: k_locations\n k_locations_costcntr\n k_locations_location \n\nwith the following query:\n\nexplain select supply from sales where target in\n(select location from locations);\nNOTICE: QUERY PLAN:\n\nSeq Scan on sales (cost=5738.60 size=116806 width=12)\n SubPlan\n -> Seq Scan on locations (cost=7.49 size=136 width=12)\n\nEXPLAIN\n\nThis results in a full table scan on the outer table\nwhere there are 150K rows. As a result, it is the\nequivalent of a full table scan on 150K (sales) * \n100 (locations) rows = 15,000,000 row table scan.\n\nOracle 8.0.5.00 for Linux Early Adaptor Program \ncorrectly uses the index on the same two tables\nresulting on a sequential access on the subselected\ntable (locations) and an index look-up on the \nselect table (sales), according to EXPLAIN PLAN.\n\nDue to the failure of PostgreSQL 6.4 to correctly\nuse indexes with subselects, it makes subselects\npractically useless.\n\nAre there any patches available to fix this problem?\n\nThanks for any information,\n\nMarcus Mascari\n([email protected])\n\n\n\n\n_________________________________________________________\nDO YOU YAHOO!?\nGet your free @yahoo.com address at http://mail.yahoo.com\n\n",
"msg_date": "Mon, 19 Oct 1998 10:39:11 -0700 (PDT)",
"msg_from": "Marcus Mascari <[email protected]>",
"msg_from_op": true,
"msg_subject": "subselect not using indexes under 6.4 - Please help"
}
] |
[
{
"msg_contents": "> \n> \tWith the talk of PostgresSQL as a DB engine for\n> 4.0 you will need to fetch a snapshot not a release\n> AND install BISON before you start the build.\n> \n> \tThe official release does not know about\n> BSD/OS 4.0 yet. A snapshot from 10-18-1998 does \n> know about 4.0 and does build. I don't know what\n> may or not be broken or enhanced.\n\n[BCC to PostgreSQL hackers list, as a reminder.]\n\nWe are expecting a 6.4 final release on November 1, and it will support\nBSD/OS 4.0 out-of-the-box, with full ELF/dynamic loader/shared library\nsupport.\n\nBison will not be required, but is helpful. Our parser grammer is too\nlarge for the BSDI yacc.\n\nAs part of the distribution, we generate a gram.c file from gram.y. It\nworks just fine with _no_ yacc/bison.\n\nHowever, there was a datestamp problem with 6.3.2, and the gram.c file\ncame out older than gram.y, so Make thought you needed yacc/bison, and\nBSD/OS yacc could not handle the size.\n\nWe supplied a patch from our ftp server ftp.postgresql.org to fix this\nby modifying the timestamp on gram.c.\n\n6.4 will have this fixed.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 19 Oct 1998 13:46:18 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Postgres for 4.0"
}
] |
[
{
"msg_contents": "Ahem:\n\nC:\\>type x\nBEGIN AUTH REQUEST\n /usr/local/cvsroot\n anoncvs\n A:0Z,I dZq\n END AUTH REQUEST\n\nC:\\>nc postgresql.org 2401 < x\nI LOVE YOU\nE Fatal error, aborting.\nerror 0 : no such user\n\nSay what? Who ate anoncvs?\n\nTaral\n",
"msg_date": "Mon, 19 Oct 1998 14:58:11 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Whodunit? Who ate anoncvs?"
},
{
"msg_contents": "\nOn 19-Oct-98 Taral wrote:\n> Ahem:\n> \n> C:\\>type x\n> BEGIN AUTH REQUEST\n> /usr/local/cvsroot\n> anoncvs\n> A:0Z,I dZq\n> END AUTH REQUEST\n> \n> C:\\>nc postgresql.org 2401 < x\n> I LOVE YOU\n> E Fatal error, aborting.\n> error 0 : no such user\n> \n> Say what? Who ate anoncvs?\n> \n> Taral\n> \n\nPerhaps it's been fixed? It's working fine for me as I type this.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Mon, 19 Oct 1998 16:11:31 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Whodunit? Who ate anoncvs?"
},
{
"msg_contents": "> > I LOVE YOU\n> > E Fatal error, aborting.\n> > error 0 : no such user\n> \n> Perhaps it's been fixed? It's working fine for me as I type this.\n> \n> Vince.\n\nNope... and I'm not the only one experiencing this problem, apparently.\n\nTaral\n",
"msg_date": "Mon, 19 Oct 1998 15:22:33 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Whodunit? Who ate anoncvs?"
}
] |
[
{
"msg_contents": "I have seen other postings about Y2K compliance of postgres but no answers. Can anyone tell me how\ncompliant Postgres is?\nThanks!\n---------\nChris Williams\nSterling Software\nRome, New York\nPhone: (315) 336-0500\nEmail: [email protected]\n\n",
"msg_date": "Mon, 19 Oct 1998 16:45:40 -0400",
"msg_from": "\"Chris Williams\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Postgres - Y2K Compliant....Yes or No"
},
{
"msg_contents": "> I have seen other postings about Y2K compliance of postgres but no \n> answers. Can anyone tell me how compliant Postgres is?\n\nYes. It is, or should be, compliant. (btw, I wrote or integrated most of\nthe date/time routines.) The only place where you might possibly run\ninto trouble is if your underlying OS has Y2K trouble in the \"get\ncurrent system time\" routine.\n\nOf course, ymmv. I've done some testing on this, and am familiar with\nmuch of the code, but more testing or possibly docs on what testing\n_has_ been done would be helpful. Anyone else have a start on this?\n\n - Tom\n",
"msg_date": "Mon, 19 Oct 1998 22:31:58 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Postgres - Y2K Compliant....Yes or No"
},
{
"msg_contents": "Thus spake Chris Williams\n> I have seen other postings about Y2K compliance of postgres but no answers. Can anyone tell me how\n> compliant Postgres is?\n\ndarcy=> select 'NOW'::timestamp;\n?column? \n----------------------\n1998-10-19 17:45:27-04\n(1 row)\n\nNow 2038 compliance is another matter. :-)\n\nP.S. I'm sure we'll have 8 byte times by then.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Mon, 19 Oct 1998 21:57:38 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Postgres - Y2K Compliant....Yes or No"
},
{
"msg_contents": "\nOn Mon, 19 Oct 1998, D'Arcy J.M. Cain wrote:\n\n> Thus spake Chris Williams\n> > I have seen other postings about Y2K compliance of postgres but no answers. Can anyone tell me how\n> > compliant Postgres is?\n> \n> darcy=> select 'NOW'::timestamp;\n> ?column? \n> ----------------------\n> 1998-10-19 17:45:27-04\n> (1 row)\n> \n> Now 2038 compliance is another matter. :-)\n> \n> P.S. I'm sure we'll have 8 byte times by then.\n\n 2038 is for 31 bit (signed int) times, if we simply go to a unsigned int\nthat will extend things for another 68 years, and break very few things.\nBy 2106, I'm sure we'll have something better to do.\n\n> -- \n> D'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\n> http://www.druid.net/darcy/ | and a sheep voting on\n> +1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n\nTom\n\n",
"msg_date": "Mon, 19 Oct 1998 20:22:16 -0700 (PDT)",
"msg_from": "Tom <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Postgres - Y2K Compliant....Yes or No"
},
{
"msg_contents": "> 2038 is for 31 bit (signed int) times, if we simply go to a unsigned \n> int that will extend things for another 68 years, and break very few \n> things. By 2106, I'm sure we'll have something better to do.\n\nHmm. I'm hoping that I'm still around in 2039 to be *really annoyed* if\nI can't store my pre-1970 birthdate in Postgres ;)\n\n - Thomas\n",
"msg_date": "Tue, 20 Oct 1998 05:07:17 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Postgres - Y2K Compliant....Yes or No"
}
] |
[
{
"msg_contents": "\nMorning...\n\n\tYes, there is a problem with CVS that I hope to have fixed by the\nend of today, if all goes well.\n\n\tDue to various requirements, I *have* to upgrade the OS on Hub,\nand for some stupid reason, the version of CVS that does the anoncvs stuff\nwon't read the FreeBSD CVS repository, so I had to backtrack :(\n\n\tOnce the upgrade is finished, I will re-install the newer cvs\nrequired for the PostgreSQL repository ...\n\n\tSorry about that :(\n\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Tue, 20 Oct 1998 07:29:27 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "anoncvs and remote CVS..."
}
] |
[
{
"msg_contents": "Hi all,\n\nI'm trying v6.4beta2 via ODBC-Access.\n\nI have the following table:\n\nCREATE TABLE attivita (\n azienda CHAR(11) NOT NULL,\n attivita CHAR(03) NOT NULL,\n specifica CHAR(02),\n cod_alternativo CHAR(10),\n conto_economico CHAR(02),\n inizio_attivita DATE,\n fine_attivita DATE,\n PRIMARY KEY (azienda,attivita,specifica,inizio_attivita)\n );\n\nI linked this table on M$-Access specifying the Access-key equal to\nPostgreSQL PRIMARY KEY to test indexes with the OR clause but\nI have the following error message:\n\n'FATAL 1: palloc failure memory exhausted'\n\nAny ideas ?\n\nThere's the PSLODBC.LOG file attached.\n\nJose'\n\n\n",
"msg_date": "Tue, 20 Oct 1998 15:52:06 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": true,
"msg_subject": "using indexes with the OR clause"
},
{
"msg_contents": "In the \"Connect Settings\" (either global or per data source) add the\nfollowing line:\n SET ksqo TO 'on'\n\nStands for \"Key Set Query Oprimizer\". It actually breaked those nasty\nOR's into UNIONS. There will eventually be a radio button for this.\nWe have be in a crunch lately at work.\n\nAre you using row versioning? If so, you may have to overload an\noperator for > on xid.\n\nLet me know how it goes.\n\n\nSferacarta Software wrote:\n\n> Hi all,\n>\n> I'm trying v6.4beta2 via ODBC-Access.\n>\n> I have the following table:\n>\n> CREATE TABLE attivita (\n> azienda CHAR(11) NOT NULL,\n> attivita CHAR(03) NOT NULL,\n> specifica CHAR(02),\n> cod_alternativo CHAR(10),\n> conto_economico CHAR(02),\n> inizio_attivita DATE,\n> fine_attivita DATE,\n> PRIMARY KEY (azienda,attivita,specifica,inizio_attivita)\n> );\n>\n> I linked this table on M$-Access specifying the Access-key equal to\n> PostgreSQL PRIMARY KEY to test indexes with the OR clause but\n> I have the following error message:\n>\n> 'FATAL 1: palloc failure memory exhausted'\n>\n> Any ideas ?\n>\n> There's the PSLODBC.LOG file attached.\n>\n> Jose'\n\n\n\n",
"msg_date": "Tue, 20 Oct 1998 11:34:38 -0400",
"msg_from": "David Hartwig <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [INTERFACES] using indexes with the OR clause"
},
{
"msg_contents": "\n\nJose' Soares wrote:\n\n> David Hartwig wrote:\n> >\n> > In the \"Connect Settings\" (either global or per data source) add the\n> > following line:\n> > SET ksqo TO 'on'\n> >\n> > Stands for \"Key Set Query Oprimizer\". It actually breaked those nasty\n> > OR's into UNIONS. There will eventually be a radio button for this.\n> > We have be in a crunch lately at work.\n> >\n> Yes, this works well now.\n>\n> > Are you using row versioning? If so, you may have to overload an\n> > operator for > on xid.\n>\n> Do you mean < I think, because my log says ERROR: Unable to find an\n> ordering operator '<' for type xid\n> Any way. I tried to create the operators =, < and > but I have some\n> troubles to do this.\n>\n> I modified xidint4.c and xidint4.sql, the sources that you sent me some\n> time ago, as follow:\n>\n> ---------------xidint4.c-------------\n> /* Insight Distribution Systems - System V - Apr 1998i\n> static char accntnum_c[] = \"@(#)accntnum.c 1.1\n> /sccs/sql/extend/s.accntnum.\n> */\n> #include <stdio.h> /* for sprintf() */\n> #include <string.h>\n> #include \"postgres.h\"\n> #include \"utils/palloc.h\"\n>\n> bool xidint4_eq(int32 arg1, int32 arg2);\n> bool xidint4_gt(int32 arg1, int32 arg2);\n> bool xidint4_lt(int32 arg1, int32 arg2);\n>\n> bool xidint4_eq(int32 arg1, int32 arg2)\n> {\n> return (arg1 == arg2);\n> }\n>\n> bool xidint4_gt(int32 arg1, int32 arg2)\n> {\n> return (arg1 > arg2);\n> }\n>\n> bool xidint4_lt(int32 arg1, int32 arg2)\n> {\n> return (arg1 < arg2);\n> }\n>\n> --------------------xidint4.sql---------------------------\n> create function xidint4_eq(xid,int4)\n> returns bool\n> as '/usr/local/pgsql/lib/contrib/xidint4.so'\n> language 'c';\n>\n> create function xidint4_gt(xid,int4)\n> returns bool\n> as '/usr/local/pgsql/lib/contrib/xidint4.so'\n> language 'c';\n>\n> create function xidint4_lt(xid,int4)\n> returns bool\n> as '/usr/local/pgsql/lib/contrib/xidint4.so'\n> language 'c';\n>\n> create operator = (\n> leftarg=xid,\n> rightarg=int4,\n> procedure=xidint4_eq,\n> commutator='=',\n> negator='<>',\n> restrict=eqsel,\n> join=eqjoinsel\n> );\n>\n> create operator < (\n> leftarg=xid,\n> rightarg=int4,\n> procedure=xidint4_lt,\n> commutator='<',\n> negator='>',\n> restrict=intltsel,\n> join=intltjoinsel\n> );\n>\n> create operator > (\n> leftarg=xid,\n> rightarg=int4,\n> procedure=xidint4_gt,\n> commutator='>',\n> negator='<',\n> restrict=intgtsel,\n> join=intgtjoinsel\n> );\n>\n> This script gives me this message for every operator it creates,\n> NOTICE: buffer leak [392] detected in BufferPoolCheckLeak()\n> CREATE\n>\n> but at end seems that it works, I can query a table as:\n>\n> select xmin from attivita where xmin = 92017;\n> select xmin from attivita where xmin > 92016;\n> select xmin from attivita where xmin < 92018;\n>\n> But psqlodbc.log has still the message:\n>\n> ERROR: Unable to find an ordering operator '<' for type xid\n>\n> Thanks David for your help.\n>\n> Jose'\n\nHmmm... Sound a bit strange. Lets keep an eye on it. Anyway, I have a\nminimalist patch which is confined to SQL which accomplishes the same thing.\nI just overload the int4 functions. Works well. If you use this patch, be\nsure to DROP the other XID operators and functions that you just created to\nremove any ambiguity.\n\n++++++++++++++++++++++++++++++++++++++\n\ncreate function int4eq(xid,int4)\n returns bool\n as ''\n language 'internal';\n\ncreate operator = (\n leftarg=xid,\n rightarg=int4,\n procedure=int4eq,\n commutator='=',\n negator='<>',\n restrict=eqsel,\n join=eqjoinsel\n );\n\ncreate function int4lt(xid,xid)\n returns bool\n as ''\n language 'internal';\n\ncreate function int4lt(xid,xid)\n returns bool\n as ''\n language 'internal';\n\ncreate operator < (\n leftarg=xid,\n rightarg=xid,\n procedure=int4lt,\n commutator='=',\n negator='<>',\n restrict=eqsel,\n join=eqjoinsel\n );\n\n\n\n\n",
"msg_date": "Wed, 21 Oct 1998 08:39:53 -0400",
"msg_from": "David Hartwig <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: [INTERFACES] using indexes with the OR clause"
}
] |
[
{
"msg_contents": "Hi,\n\n while writing some documentation for the rule system I found\n this little bug. Causes crashing backend.\n\n Please apply.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\ndiff -cr rewrite.orig/rewriteHandler.c rewrite/rewriteHandler.c\n*** rewrite.orig/rewriteHandler.c\tTue Oct 20 18:59:24 1998\n--- rewrite/rewriteHandler.c\tTue Oct 20 18:55:19 1998\n***************\n*** 1789,1799 ****\n \t\t\t\t\t\treturn;\n \t\t\t\t\t}\n \n \t\t\t\t\tif (var->varlevelsup > 0 &&\n \t\t\t\t\t\t\tnodeTag(exp) == T_Var) {\n- \t\t\t\t\t\texp = copyObject(exp);\n \t\t\t\t\t\t((Var *)exp)->varlevelsup = var->varlevelsup;\n! \t\t\t\t\t}\n \t\t\t\t\t*nodePtr = exp;\n \t\t\t\t\t*modified = TRUE;\n \t\t\t\t}\n--- 1789,1799 ----\n \t\t\t\t\t\treturn;\n \t\t\t\t\t}\n \n+ \t\t\t\t\texp = copyObject(exp);\n \t\t\t\t\tif (var->varlevelsup > 0 &&\n \t\t\t\t\t\t\tnodeTag(exp) == T_Var) {\n \t\t\t\t\t\t((Var *)exp)->varlevelsup = var->varlevelsup;\n! \t\t\t\t\t} \n \t\t\t\t\t*nodePtr = exp;\n \t\t\t\t\t*modified = TRUE;\n \t\t\t\t}\ndiff -cr rewrite.orig/rewriteManip.c rewrite/rewriteManip.c\n*** rewrite.orig/rewriteManip.c\tTue Oct 20 18:59:24 1998\n--- rewrite/rewriteManip.c\tTue Oct 20 18:58:18 1998\n***************\n*** 418,431 ****\n \t\t\t\t\t{\n \t\t\t\t\t\tif (info->event == CMD_UPDATE)\n \t\t\t\t\t\t{\n! \t\t\t\t\t\t\t((Var *) node)->varno = info->current_varno;\n! \t\t\t\t\t\t\t((Var *) node)->varnoold = info->current_varno;\n \t\t\t\t\t\t}\n \t\t\t\t\t\telse\n \t\t\t\t\t\t\t*nodePtr = make_null(((Var *) node)->vartype);\n \t\t\t\t\t}\n \t\t\t\t\telse\n! \t\t\t\t\t\t*nodePtr = n;\n \t\t\t\t}\n \t\t\t\tbreak;\n \t\t\t}\n--- 418,432 ----\n \t\t\t\t\t{\n \t\t\t\t\t\tif (info->event == CMD_UPDATE)\n \t\t\t\t\t\t{\n! \t\t\t\t\t\t\t*nodePtr = n = copyObject(node);\n! \t\t\t\t\t\t\t((Var *) n)->varno = info->current_varno;\n! \t\t\t\t\t\t\t((Var *) n)->varnoold = info->current_varno;\n \t\t\t\t\t\t}\n \t\t\t\t\t\telse\n \t\t\t\t\t\t\t*nodePtr = make_null(((Var *) node)->vartype);\n \t\t\t\t\t}\n \t\t\t\t\telse\n! \t\t\t\t\t\t*nodePtr = copyObject(n);\n \t\t\t\t}\n \t\t\t\tbreak;\n \t\t\t}\n",
"msg_date": "Tue, 20 Oct 1998 19:02:17 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Bugfix for rule system"
},
{
"msg_contents": "Applied. CVS is working for me.\n\n> Hi,\n> \n> while writing some documentation for the rule system I found\n> this little bug. Causes crashing backend.\n> \n> Please apply.\n> \n> \n> Jan\n> \n> --\n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n> \n> \n> diff -cr rewrite.orig/rewriteHandler.c rewrite/rewriteHandler.c\n> *** rewrite.orig/rewriteHandler.c\tTue Oct 20 18:59:24 1998\n> --- rewrite/rewriteHandler.c\tTue Oct 20 18:55:19 1998\n> ***************\n> *** 1789,1799 ****\n> \t\t\t\t\t\treturn;\n> \t\t\t\t\t}\n> \n> \t\t\t\t\tif (var->varlevelsup > 0 &&\n> \t\t\t\t\t\t\tnodeTag(exp) == T_Var) {\n> - \t\t\t\t\t\texp = copyObject(exp);\n> \t\t\t\t\t\t((Var *)exp)->varlevelsup = var->varlevelsup;\n> ! \t\t\t\t\t}\n> \t\t\t\t\t*nodePtr = exp;\n> \t\t\t\t\t*modified = TRUE;\n> \t\t\t\t}\n> --- 1789,1799 ----\n> \t\t\t\t\t\treturn;\n> \t\t\t\t\t}\n> \n> + \t\t\t\t\texp = copyObject(exp);\n> \t\t\t\t\tif (var->varlevelsup > 0 &&\n> \t\t\t\t\t\t\tnodeTag(exp) == T_Var) {\n> \t\t\t\t\t\t((Var *)exp)->varlevelsup = var->varlevelsup;\n> ! \t\t\t\t\t} \n> \t\t\t\t\t*nodePtr = exp;\n> \t\t\t\t\t*modified = TRUE;\n> \t\t\t\t}\n> diff -cr rewrite.orig/rewriteManip.c rewrite/rewriteManip.c\n> *** rewrite.orig/rewriteManip.c\tTue Oct 20 18:59:24 1998\n> --- rewrite/rewriteManip.c\tTue Oct 20 18:58:18 1998\n> ***************\n> *** 418,431 ****\n> \t\t\t\t\t{\n> \t\t\t\t\t\tif (info->event == CMD_UPDATE)\n> \t\t\t\t\t\t{\n> ! \t\t\t\t\t\t\t((Var *) node)->varno = info->current_varno;\n> ! \t\t\t\t\t\t\t((Var *) node)->varnoold = info->current_varno;\n> \t\t\t\t\t\t}\n> \t\t\t\t\t\telse\n> \t\t\t\t\t\t\t*nodePtr = make_null(((Var *) node)->vartype);\n> \t\t\t\t\t}\n> \t\t\t\t\telse\n> ! \t\t\t\t\t\t*nodePtr = n;\n> \t\t\t\t}\n> \t\t\t\tbreak;\n> \t\t\t}\n> --- 418,432 ----\n> \t\t\t\t\t{\n> \t\t\t\t\t\tif (info->event == CMD_UPDATE)\n> \t\t\t\t\t\t{\n> ! \t\t\t\t\t\t\t*nodePtr = n = copyObject(node);\n> ! \t\t\t\t\t\t\t((Var *) n)->varno = info->current_varno;\n> ! \t\t\t\t\t\t\t((Var *) n)->varnoold = info->current_varno;\n> \t\t\t\t\t\t}\n> \t\t\t\t\t\telse\n> \t\t\t\t\t\t\t*nodePtr = make_null(((Var *) node)->vartype);\n> \t\t\t\t\t}\n> \t\t\t\t\telse\n> ! \t\t\t\t\t\t*nodePtr = copyObject(n);\n> \t\t\t\t}\n> \t\t\t\tbreak;\n> \t\t\t}\n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 13:22:30 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Bugfix for rule system"
}
] |
[
{
"msg_contents": "I assume this item is done:\n\n\tfix perl/tcl compile/library issues\n\n---------------------------------------------------------------------------\n\nAdditions\n---------\nnew functoins and test INET address type(Tom Helbekkmo,D'Arcy)\nCREATE TABLE test (x text, s serial) fails if no database creation permission\nregression test all platforms\n\nSerious Items\n------------\nchange pg args for platforms that don't support argv changes\n\t(setproctitle()?, sendmail hack?)\n\nDocs\n----\ngenerate html/postscript documentation\nmake sure all changes are documented properly\n\nMinor items\n-----------\ncnf-ify still can exhaust memory, make SET KSQO more generic\npermissions on indexes: what do they do? should it be prevented?\nallow multiple generic operators in expressions without the use of parentheses\ndocument/trigger/rule so changes to pg_shadow create pg_pwd\nlarge objects orphanage\nimprove group handling\nimprove PRIMARY KEY handling\ngenerate postmaster pid file and remove flock/fcntl lock code\nadd ability to specifiy location of lock/socket files\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 15:19:16 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Open 6.4 issues"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> I assume this item is done:\n> \tfix perl/tcl compile/library issues\n\nUnless new complaints come down the pike, I think it is done.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 20 Oct 1998 16:57:20 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 issues "
}
] |
[
{
"msg_contents": "Has anyone tested pg_upgrade?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 19:22:46 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "pg_upgrade"
}
] |
[
{
"msg_contents": "as most of you know by now, jon postel, the original IANA, died last week.\ni'm sorry that i've been so flakey in my correspondance here but it's been\na rough time for those of us who knew him and/or who were helping him turn\nIANA into a new nonprofit corporation called ICANN. see http://www.iana.org/\nfor more details.\n\nthe original \"cidr\" type was written as part of a prototype registry to let\nIANA take back some of the duties it had outsourced to InterNIC. so while\nCOM, NET, and ORG are likely to continue to be operated by NSI/InterNIC for\nthe next few years, and while the networks and ASNs are being handed out via\nregional registries, there is a \"top level\" registry function that IANA, and\nnow ICANN, has to provide directly.\n\ni wanted this to be done with postgres, and i am still working toward that.\nhere are some brief notes on the overall database structure i envisioned, in\ncase anyone is wondering about the application \"cidr\" was written for. note\nthat the type will at present be called \"inet\" in pgsql 6.4.\n\ni would be very happy if someone with access to the pgsql 6.4 sources would\nadd a comment to the top of the file which implements the CIDR/INET type,\nto the effect of /* JBP RIP 16Oct98 */. thanks.\n\n---------------------------------------------------------------- countries\n\nDROP TABLE countries;\nCREATE TABLE countries (\n\tcode\tCHAR(2)\t\tNOT NULL,\t-- US\n\tname\tVARCHAR(25)\tNOT NULL\t-- United States\n);\nCREATE UNIQUE INDEX country_codes ON countries ( code );\nINSERT INTO countries VALUES ('US', 'United States');\nINSERT INTO countries VALUES ('CA', 'Canada');\nINSERT INTO countries VALUES ('MX', 'Mexico');\nINSERT INTO countries VALUES ('UK', 'United Kingdom');\nINSERT INTO countries VALUES ('SE', 'Sweden');\n-- XXX more needed here\nGRANT all ON countries TO www;\n\n---------------------------------------------------------------- contacts\n\nDROP TABLE contact_types;\nCREATE TABLE contact_types (\n\tcode\tCHAR(1)\t\tNOT NULL,\t-- T\n\tdescr\tVARCHAR(25)\tNOT NULL\t-- Technical\n);\nCREATE UNIQUE INDEX contact_codes ON contact_types ( code );\nINSERT INTO contact_types VALUES ('B', 'Billing');\nINSERT INTO contact_types VALUES ('A', 'Administrative');\nINSERT INTO contact_types VALUES ('R', 'Registry');\nINSERT INTO contact_types VALUES ('T', 'Technical');\n\nDROP TABLE notify_types;\nCREATE TABLE notify_types (\n\tcode\tCHAR(1)\t\tNOT NULL,\t-- A\n\tdescr\tVARCHAR(25)\tNOT NULL\t-- After\n);\nCREATE UNIQUE INDEX notify_codes ON notify_types ( code );\nINSERT INTO notify_types VALUES ('B', 'Before');\nINSERT INTO notify_types VALUES ('A', 'After');\nINSERT INTO notify_types VALUES ('N', 'Never');\n\nDROP TABLE contacts;\nCREATE TABLE contacts (\n\thandle\tVARCHAR(16)\tNOT NULL,\t-- PV15\n\tname\tTEXT\t\tNOT NULL,\t-- Paul Vixie\n\temail\tVARCHAR(96)\tNOT NULL,\t-- [email protected]\n\tpmail\tTEXT\t\tNOT NULL,\t-- 950 Charter Street\n\t\t\t\t\t\t-- Redwood City, CA\n\tpcode\tVARCHAR(64)\tNOT NULL,\t-- 94062\n\tcountry\tCHAR(2)\t\tNOT NULL,\t-- US\n\tphone1\tVARCHAR(64)\tNOT NULL,\t-- +1.650.779.7001\n\tphone2\tVARCHAR(64)\t,\n\tpgpkid\tCHAR(8)\t\tNOT NULL,\t-- 8972C7C1\n\tntype\tCHAR(1)\t\tNOT NULL,\t-- notify A\n\tcomment\tTEXT\t\t,\n\twww\tVARCHAR(96)\t,\t\t-- http://www.vix.com/\n\tformat\tFLOAT\t\t,\t\t-- 1.0\n\tcreated\tDATETIME\tNOT NULL,\n\tupdated\tDATETIME\tNOT NULL\n);\nCREATE UNIQUE INDEX contact_handles ON contacts ( handle );\n-- wish there was a way to require country to match a key in countries.\nGRANT all ON contacts TO www;\n\n---------------------------------------------------------------- hosts\n\nDROP TABLE host_contacts;\nCREATE TABLE host_contacts (\n\thost\tVARCHAR(64)\tNOT NULL,\t-- gw.home.vix.com\n\tctype\tCHAR(1)\t\tNOT NULL,\t-- T\n\thandle\tVARCHAR(16)\tNOT NULL,\t-- PV15\n\tntype\tCHAR(1)\t\t\t\t-- notify\n);\nCREATE UNIQUE INDEX host_contacts_index ON host_contacts\n\t\t\t\t\t\t( host, ctype, handle );\nGRANT all ON host_contacts TO www;\n\nDROP TABLE host_addresses;\nCREATE TABLE host_addresses (\n\thost\tVARCHAR(64)\tNOT NULL,\t-- gw.home.vix.com\n\taddr\tCIDR\t\tNOT NULL\t-- 192.5.5.1/32\n);\nCREATE UNIQUE INDEX host_addresses_index ON host_addresses\n\t\t\t\t\t\t( host, addr );\n\nDROP TABLE hosts;\nCREATE TABLE hosts (\n\thost\tVARCHAR(64)\tNOT NULL,\t-- gw.home.vix.com\n\tdescr\tTEXT\t\t,\n\tcomment\tTEXT\t\t,\t\t-- see http://www.vix.com/\n\tformat\tFLOAT\t\t,\t\t-- 1.0\n\tcreated\tDATETIME\tNOT NULL,\n\tupdated\tDATETIME\tNOT NULL\n);\nCREATE UNIQUE INDEX host_names ON hosts ( host );\nGRANT all ON hosts TO www;\n\n---------------------------------------------------------------- netblocks\n\nDROP TABLE netblock_contacts;\nCREATE TABLE netblock_contacts (\n\tnet\tCIDR\t\tNOT NULL,\t-- 192.5.4/23\n\tctype\tCHAR(1)\t\tNOT NULL,\t-- T\n\thandle\tVARCHAR(16)\tNOT NULL,\t-- PV15\n\tntype\tCHAR(1)\t\t\t\t-- notify\n);\nCREATE UNIQUE INDEX netblock_contacts_index ON netblock_contacts\n\t\t\t\t\t\t( net, ctype, handle );\nGRANT all ON netblock_contacts TO www;\n\nDROP TABLE netblock_nameservers;\nCREATE TABLE netblock_nameservers (\n\tnet\tCIDR\t\tNOT NULL,\t-- 192.5.4/23\n\thost\tVARCHAR(64)\tNOT NULL\t-- gw.home.vix.com\n);\nCREATE UNIQUE INDEX netblock_nameservers_index ON netblock_nameservers\n\t\t\t\t\t\t( net, host );\nGRANT all ON netblock_nameservers TO www;\n\nDROP TABLE netblocks;\nCREATE TABLE netblocks (\n\tnet\tCIDR\t\tNOT NULL,\t-- 192.5.4/23\n\tdescr\tTEXT\t\t,\n\tcomment\tTEXT\t\t,\n\tzonekey\tTEXT\t\t,\n\twhois\tVARCHAR(64)\t,\t\t-- whois.vix.com\n\twww\tVARCHAR(96)\t,\t\t-- http://www.vix.com/\n\tformat\tFLOAT\t\t,\t\t-- 1.0\n\tcreated\tDATETIME\tNOT NULL,\n\tupdated\tDATETIME\tNOT NULL\n);\nCREATE UNIQUE INDEX netblock_nets ON netblocks ( net );\nGRANT all ON netblocks TO www;\n\n---------------------------------------------------------------- domains\n\nDROP TABLE domain_contacts;\nCREATE TABLE domain_contacts (\n\tdomain\tVARCHAR(64)\tNOT NULL,\t-- VIX.COM\n\tctype\tCHAR(1)\t\tNOT NULL,\t-- T\n\thandle\tVARCHAR(16)\tNOT NULL,\t-- PV15\n\tntype\tCHAR(1)\t\t\t\t-- notify\n);\nCREATE UNIQUE INDEX domain_contacts_index ON domain_contacts\n\t\t\t\t\t\t( domain, ctype, handle );\nGRANT all ON domain_contacts TO www;\n\nDROP TABLE domain_nameservers;\nCREATE TABLE domain_nameservers (\n\tdomain\tVARCHAR(64)\tNOT NULL,\t-- VIX.COM\n\thost\tVARCHAR(64)\tNOT NULL\t-- gw.home.vix.com\n);\nCREATE UNIQUE INDEX domain_nameservers_index ON domain_nameservers\n\t\t\t\t\t\t( domain, host );\nGRANT all ON domain_nameservers TO www;\n\nDROP TABLE domains;\nCREATE TABLE domains (\n\tdomain\tVARCHAR(64)\tNOT NULL,\t-- VIX.COM\n\tdescr\tTEXT\t\t,\n\tcomment\tTEXT\t\t,\n\tzonekey\tTEXT\t\t,\n\twhois\tVARCHAR(64)\t,\t\t-- whois.vix.com\n\twww\tVARCHAR(96)\t,\t\t-- http://www.vix.com/\n\tformat\tFLOAT\t\t,\t\t-- 1.0\n\tcreated\tDATETIME\tNOT NULL,\n\tupdated\tDATETIME\tNOT NULL\n);\nCREATE UNIQUE INDEX domain_domains ON domains ( domain );\nGRANT all ON domains TO www;\n\n---------------------------------------------------------------- asblks\n\nDROP TABLE asblk_contacts;\nCREATE TABLE asblk_contacts (\n\tasblk\tVARCHAR(64)\tNOT NULL,\t-- BARRNET-BLK\n\tctype\tCHAR(1)\t\tNOT NULL,\t-- T\n\thandle\tVARCHAR(16)\tNOT NULL,\t-- PV15\n\tntype\tCHAR(1)\t\t\t\t-- notify\n);\nCREATE UNIQUE INDEX asblk_contacts_index ON asblk_contacts\n\t\t\t\t\t\t( asblk, ctype, handle );\nGRANT all ON asblk_contacts TO www;\n\nDROP TABLE asblks;\nCREATE TABLE asblks (\n\tasblk\tVARCHAR(64)\tNOT NULL,\t-- BARRNET-BLK\n\tfirst\tINT\t\tNOT NULL,\t-- 199\n\tlast\tINT\t\tNOT NULL,\t-- 203\n\tdescr\tTEXT\t\t,\n\tcomment\tTEXT\t\t,\n\twhois\tVARCHAR(64)\t,\t\t-- whois.vix.com\n\twww\tVARCHAR(96)\t,\t\t-- http://www.vix.com/\n\tformat\tFLOAT\t\t,\t\t-- 1.0\n\tcreated\tDATETIME\t,\n\tupdated\tDATETIME\tNOT NULL\n);\nCREATE UNIQUE INDEX asblk_names ON asblks ( asblk );\nGRANT all ON asblks TO www;\n\n---------------------------------------------------------------- :EOF:\n",
"msg_date": "Tue, 20 Oct 1998 17:46:55 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": true,
"msg_subject": "CIDR/INET type and IANA/ICANN"
},
{
"msg_contents": "> as most of you know by now, jon postel, the original IANA, died last week.\n> i'm sorry that i've been so flakey in my correspondance here but it's been\n> a rough time for those of us who knew him and/or who were helping him turn\n> IANA into a new nonprofit corporation called ICANN. see http://www.iana.org/\n> for more details.\n> \n> the original \"cidr\" type was written as part of a prototype registry to let\n> IANA take back some of the duties it had outsourced to InterNIC. so while\n> COM, NET, and ORG are likely to continue to be operated by NSI/InterNIC for\n> the next few years, and while the networks and ASNs are being handed out via\n> regional registries, there is a \"top level\" registry function that IANA, and\n> now ICANN, has to provide directly.\n> \n> i wanted this to be done with postgres, and i am still working toward that.\n> here are some brief notes on the overall database structure i envisioned, in\n> case anyone is wondering about the application \"cidr\" was written for. note\n> that the type will at present be called \"inet\" in pgsql 6.4.\n> \n> i would be very happy if someone with access to the pgsql 6.4 sources would\n> add a comment to the top of the file which implements the CIDR/INET type,\n> to the effect of /* JBP RIP 16Oct98 */. thanks.\n\nAdded. I put this at the top of the inet.c file, and next to your name\non the TODO list. It is on the web site TODO list now. Hope it is OK. \nI have spelled out his full name:\n\n * Paul A Vixie <[email protected]> (Jon Postel RIP 16 Oct 1998)\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 23:11:32 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CIDR/INET type and IANA/ICANN"
},
{
"msg_contents": "On Tue, 20 Oct 1998, Paul A Vixie wrote:\n\nCREATE FUNCTION check_primary_key ()\n\tRETURNS opaque\n\tAS '/opt/PGpgsql/modules/refint.so'\t-- replace with real path\n\tLANGUAGE 'c'\n;\n\nCREATE FUNCTION check_foreign_key ()\n\tRETURNS opaque\n\tAS '/opt/PGpgsql/modules/refint.so'\t-- replace with real path\n\tLANGUAGE 'c'\n;\n\n> CREATE TABLE countries (\n> \tcode\tCHAR(2)\t\tNOT NULL,\t-- US\n> \tname\tVARCHAR(25)\tNOT NULL,\t-- United States\n\n\tPRIMARY KEY (code)\n\n> );\n> CREATE UNIQUE INDEX country_codes ON countries ( code );\n> INSERT INTO countries VALUES ('US', 'United States');\n> INSERT INTO countries VALUES ('CA', 'Canada');\n> INSERT INTO countries VALUES ('MX', 'Mexico');\n> INSERT INTO countries VALUES ('UK', 'United Kingdom');\n> INSERT INTO countries VALUES ('SE', 'Sweden');\n> -- XXX more needed here\n> GRANT all ON countries TO www;\n\n[snip]\n\n> DROP TABLE contacts;\n> CREATE TABLE contacts (\n> \thandle\tVARCHAR(16)\tNOT NULL,\t-- PV15\n> \tname\tTEXT\t\tNOT NULL,\t-- Paul Vixie\n> \temail\tVARCHAR(96)\tNOT NULL,\t-- [email protected]\n> \tpmail\tTEXT\t\tNOT NULL,\t-- 950 Charter Street\n> \t\t\t\t\t\t-- Redwood City, CA\n> \tpcode\tVARCHAR(64)\tNOT NULL,\t-- 94062\n> \tcountry\tCHAR(2)\t\tNOT NULL,\t-- US\n> \tphone1\tVARCHAR(64)\tNOT NULL,\t-- +1.650.779.7001\n> \tphone2\tVARCHAR(64)\t,\n> \tpgpkid\tCHAR(8)\t\tNOT NULL,\t-- 8972C7C1\n> \tntype\tCHAR(1)\t\tNOT NULL,\t-- notify A\n> \tcomment\tTEXT\t\t,\n> \twww\tVARCHAR(96)\t,\t\t-- http://www.vix.com/\n> \tformat\tFLOAT\t\t,\t\t-- 1.0\n> \tcreated\tDATETIME\tNOT NULL,\n> \tupdated\tDATETIME\tNOT NULL,\n\n\t-- note: while postgresql does not support 'foreign key'\n\t-- statements, we include it here just to remind ourselves\n\t-- that we are using the refint.so calls to do the same thing.\n\tFOREIGN KEY (country) REFERENCES countries\n\t\tON DELETE CASCADE\n\t\tON UPDATE CASCADE\n> );\n> CREATE UNIQUE INDEX contact_handles ON contacts ( handle );\n> -- wish there was a way to require country to match a key in countries.\n> GRANT all ON contacts TO www;\n\nCREATE TRIGGER t_countries_country\n\tBEFORE DELETE OR UPDATE ON countries\n\tFOR EACH ROW\n\tEXECUTE PROCEDURE\n\tcheck_foreign_key(1, 'cascade', 'code',\n\t\t'contacts', 'countries');\n\nCREATE TRIGGER t_contacts_countries\n\tBEFORE INSERT OR UPDATE ON contacts\n\tFOR EACH ROW\n\tEXECUTE PROCEDURE\n\tcheck_primary_key('countries', 'countries', 'code');\n\n[snip]\n\nLook in contrib for the refint.c file and examples.\n\n-- \n| Matthew N. Dodd | 78 280Z | 75 164E | 84 245DL | FreeBSD/NetBSD/Sprite/VMS |\n| [email protected] | This Space For Rent | ix86,sparc,m68k,pmax,vax |\n| http://www.jurai.net/~winter | Are you k-rad elite enough for my webpage? |\n\n",
"msg_date": "Wed, 21 Oct 1998 12:33:06 -0400 (EDT)",
"msg_from": "\"Matthew N. Dodd\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CIDR/INET type and IANA/ICANN"
},
{
"msg_contents": "> Added. I put this at the top of the inet.c file, and next to your name\n> on the TODO list. It is on the web site TODO list now. Hope it is OK. \n> I have spelled out his full name:\n> \n> * Paul A Vixie <[email protected]> (Jon Postel RIP 16 Oct 1998)\n\nif it's all the same to you i'd like it to stand alone, unrelated to me.\n\n",
"msg_date": "Wed, 21 Oct 1998 09:42:15 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] CIDR/INET type and IANA/ICANN "
},
{
"msg_contents": "On Wed, 21 Oct 1998, Paul A Vixie wrote:\n\n> > Added. I put this at the top of the inet.c file, and next to your name\n> > on the TODO list. It is on the web site TODO list now. Hope it is OK. \n> > I have spelled out his full name:\n> > \n> > * Paul A Vixie <[email protected]> (Jon Postel RIP 16 Oct 1998)\n> \n> if it's all the same to you i'd like it to stand alone, unrelated to me.\n\n\tOdd question, but can we put some sort of \"reference\" in\nthere...if I hadn't read your original email, I would have looked at that\nand wonder what this was supposed to refer to. Maybe a reference to IANA?\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Wed, 21 Oct 1998 14:52:20 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CIDR/INET type and IANA/ICANN "
},
{
"msg_contents": "> > if it's all the same to you i'd like it to stand alone, unrelated to me.\n> \n> \tOdd question, but can we put some sort of \"reference\" in\n> there...if I hadn't read your original email, I would have looked at that\n> and wonder what this was supposed to refer to. Maybe a reference to IANA?\n\nhe was a man, not an organization, and his footprint is much larger than IANA.\n",
"msg_date": "Wed, 21 Oct 1998 11:59:53 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] CIDR/INET type and IANA/ICANN "
},
{
"msg_contents": "Thus spake Marc G. Fournier\n> > > * Paul A Vixie <[email protected]> (Jon Postel RIP 16 Oct 1998)\n> > if it's all the same to you i'd like it to stand alone, unrelated to me.\n> \n> \tOdd question, but can we put some sort of \"reference\" in\n> there...if I hadn't read your original email, I would have looked at that\n> and wonder what this was supposed to refer to. Maybe a reference to IANA?\n\nOr his web page at http://www.isi.edu/~postel/ perhaps?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Wed, 21 Oct 1998 15:39:41 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CIDR/INET type and IANA/ICANN"
},
{
"msg_contents": "subscribe\n\n\n",
"msg_date": "Thu, 22 Oct 1998 16:12:38 -0400 (EDT)",
"msg_from": "Dragana Obradovic <[email protected]>",
"msg_from_op": false,
"msg_subject": "None"
}
] |
[
{
"msg_contents": "Hi. I had ftp'd in all the files for the new ODBC code in the main\nPostgres tree. I forgot to change the protection on the derived\nodbc/configure file, so it is checked in without execute permission.\n\nDo you have any experience fixing this kind of problem? Simply checking\nin a new version with the right permissions doesn't seem to fix it. The\ntwo alternatives I'm considering are:\n - \"cvs delete\" then \"cvs add\" the file with the right permissions\n - going into the cvs tree itself and brute-force change it\n\nSuggestions? If you know what to do, and want to do it, feel free to fix\nit. I'm inclined to try the brute-force method, but remote CVSup trees\n(like mine) may need the same kludge fixup :O\n\n - Tom\n",
"msg_date": "Wed, 21 Oct 1998 02:53:01 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Protections problem on CVS tree"
},
{
"msg_contents": "> Hi. I had ftp'd in all the files for the new ODBC code in the main\n> Postgres tree. I forgot to change the protection on the derived\n> odbc/configure file, so it is checked in without execute permission.\n> \n> Do you have any experience fixing this kind of problem? Simply checking\n> in a new version with the right permissions doesn't seem to fix it. The\n> two alternatives I'm considering are:\n> - \"cvs delete\" then \"cvs add\" the file with the right permissions\n> - going into the cvs tree itself and brute-force change it\n> \n> Suggestions? If you know what to do, and want to do it, feel free to fix\n> it. I'm inclined to try the brute-force method, but remote CVSup trees\n> (like mine) may need the same kludge fixup :O\n\nGo to $CVSROOT/pgsql, and fiddle wit hthe protections of the base files,\nno?\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 20 Oct 1998 22:56:33 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Protections problem on CVS tree"
},
{
"msg_contents": "On Wed, 21 Oct 1998, Thomas G. Lockhart wrote:\n\n> Hi. I had ftp'd in all the files for the new ODBC code in the main\n> Postgres tree. I forgot to change the protection on the derived\n> odbc/configure file, so it is checked in without execute permission.\n> \n> Do you have any experience fixing this kind of problem? Simply checking\n> in a new version with the right permissions doesn't seem to fix it. The\n> two alternatives I'm considering are:\n> - \"cvs delete\" then \"cvs add\" the file with the right permissions\n> - going into the cvs tree itself and brute-force change it\n> \n> Suggestions? If you know what to do, and want to do it, feel free to fix\n> it. I'm inclined to try the brute-force method, but remote CVSup trees\n> (like mine) may need the same kludge fixup :O\n\nhub> ls -lt\ntotal 967\n-r--r--r-- 1 thomas pgsql 11086 Oct 21 01:39 GNUmakefile.in,v\n-r-xr-xr-x 1 thomas pgsql 65482 Oct 21 01:23 configure,v\n\n\npermissions appear to be fine in the repository itself...and I just\nchecked it out:\n\n-rw-r--r-- 1 scrappy wheel 8844 Oct 21 14:46 configure.in\n-rwxr-xr-x 1 scrappy wheel 60983 Oct 21 14:46 configure\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Wed, 21 Oct 1998 14:49:41 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Protections problem on CVS tree"
},
{
"msg_contents": "> permissions appear to be fine in the repository itself...and I just\n> checked it out:\n\nYeah, I went ahead and tried the brute-force fix (at Bruce's suggestion,\nso problems are all his fault :), apparently before you looked at it.\nSeems to be behaving fine.\n\n - Tom\n",
"msg_date": "Thu, 22 Oct 1998 04:45:56 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Protections problem on CVS tree"
}
] |
[
{
"msg_contents": "Ok, I have posted the patches for the inet type functions and Bruce has\napplied it. The functions added are inet_netmask(), inet_broadcast() and\ninet_masklen(). The other ones I mentioned before don't apply to\nthis type.\n\nNext issue is the names given to the functions in the user API. At\nthe moment they just have the same names as the underlying functions\nbut I think they should be netmask(), broadcast() and masklen().\nDoes anyone disagree?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Tue, 20 Oct 1998 22:59:29 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: inet/cidr/bind"
}
] |
[
{
"msg_contents": "I've posted a tarball of the new standalone psqlODBC code at\n\n ftp://postgresql.org/pub/patches/psqlodbc-025-6.4beta.tar.gz\n\nalong with a tarball of the html Programmer's Guide which contains a\nchapter on installation and usage instructions for psqlODBC. The docs\ninclude updates from Gerald, Robert, and Cary. btw, the code tarball is\nidentical to that currently in the main Postgres cvs tree afaik.\n\nI've also put new tarballs of beta versions of html docs for the Admin\nGuide, the User's Guide, and the \"integrated docs\" in the same area.\n\nOh, for background: this tarball will allow one to use Applix Data\n(4.4.1) with the Postgres RDBMS. It is based on the current psqlODBC\ndistribution at www.insightdist.com and at postgresql.org, but has minor\ncode and major configure changes to allow building on Unix systems.\n\nLet me know how it goes! I'm out of town tomorrow, but will be back\nThursday. Cary O'Brien has been providing info and user-support for the\nApplix Data ODBC interface for PostgreSQL, and will likely be posting\nmore info to help get folks started.\n\n - Tom\n",
"msg_date": "Wed, 21 Oct 1998 06:07:39 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "New docs and ODBC for Unix"
},
{
"msg_contents": "Thomas G. Lockhart wrote:\n> \n> I've posted a tarball of the new standalone psqlODBC code at\n> \n> ftp://postgresql.org/pub/patches/psqlodbc-025-6.4beta.tar.gz\n\nI don't mean to nitpick Thomas, heck I'm the one that suggested\nyou rename the distribution to temporarily distinguish it from\nthe Insightdist driver. But I'm afraid the above name might\n\"scare\" off people who think that it is only for PostgreSQL v6.4.\n\nIf its not too much trouble can you just drop the 6.4(leave in\nbeta of course).\n-- \nGerald Gryschuk\[email protected]\nMidNightOil Consulting - \"We burn the midnight oil so you don't have\nto.\"\n",
"msg_date": "Wed, 21 Oct 1998 22:45:25 -0600",
"msg_from": "Gerald Gryschuk <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [INTERFACES] New docs and ODBC for Unix"
},
{
"msg_contents": "> > I've posted a tarball of the new standalone psqlODBC code at\n> > ftp://postgresql.org/pub/patches/psqlodbc-025-6.4beta.tar.gz\n> ... I'm afraid the above name might\n> \"scare\" off people who think that it is only for PostgreSQL v6.4.\n> If its not too much trouble can you just drop the 6.4 (leave in\n> beta of course).\n\n*sigh*\n\nThe only thing I'm certain of is that this code is an exact copy of the\ncode currently in the v6.4beta cvs source tree. It isn't\npsqlodbc-025beta since it is out _after_ Byron's v025 official release.\nDon't know what Byron's next release will be called, and don't know for\ncertain that this source code will be identical to what Byron would call\nhis next beta. We're not quite to that point yet.\n\nPerhaps there has been enough discussion of this on the lists that there\nwon't be confusion for the next couple of weeks, after which v6.4 will\nbe released and Byron will probably have a new WIN32-verified zipball.\n\nIn the meantime, to be clear:\n\nThe tarball on the postgres ftp site can be used for a Unix client to\ntalk to any existing v6.2, v6.3, and v6.4beta Postgres installation with\nno other files or tarballs required. It's been tested under Linux, and\nmight have a prayer for FreeBSD and Solaris. Not certain about other\nclients or about pre-v6.2 servers, but perhaps\nhttp://www.insightdist.com/ has some mention about the server versions.\nAlmost certainly the source code itself says which versions it will play\nwith. Other clients _should_ work, they just haven't been tested and the\ninstallation info may need to be tweaked.\n\nThe zipball on the insightdist ftp site can be used for a WIN32 client\nto access the same set of server versions as above.\n\nThe two \"codeballs\" will likely be remerged soon.\n\n - Tom\n",
"msg_date": "Thu, 22 Oct 1998 05:59:32 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [INTERFACES] New docs and ODBC for Unix"
},
{
"msg_contents": "Thomas G. Lockhart wrote:\n> \n> > > I've posted a tarball of the new standalone psqlODBC code at\n> > > ftp://postgresql.org/pub/patches/psqlodbc-025-6.4beta.tar.gz\n> > ... I'm afraid the above name might\n> > \"scare\" off people who think that it is only for PostgreSQL v6.4.\n> > If its not too much trouble can you just drop the 6.4 (leave in\n> > beta of course).\n> \n> *sigh*\n\nSorry Thomas, I know how hard your working, and like you say this\nwon't be out there that long before a new version number change\nshould clear things up. \n-- \nGerald Gryschuk([email protected])\nProgrammer Analyst\nSaskatoon Cancer Centre\n((306)655-2746)\n",
"msg_date": "Thu, 22 Oct 1998 09:07:22 -0600",
"msg_from": "Gerald Gryschuk <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [INTERFACES] New docs and ODBC for Unix"
}
] |
[
{
"msg_contents": "I have recieved the files from D'Arcy for new INET and CIDR types. INET\nis use for hosts, while CIDR for networks. No duplicate code. CIDR\nbehaves the same as INET, except for a different input function. The\nsystem catalogs take care of the mapping of CIDR to INET functions.\n\nIt will require an initdb for beta users moving to 6.4 final. Is this\nOK? If so, I will apply it tomorrow. If not, the type will have to be\ndisabled for 6.4.\n\nThe big question is how many people have _live_ data in 6.4 beta\ndatabases, who don't want to initdb for 6.4 final, and can't use\npg_upgrade to move to 6.4 final.\n\nIf I don't hear any objections, I will do it some time around noon\ntomorrow, EST.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 02:14:02 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "New INET and CIDR types"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> I have recieved the files from D'Arcy for new INET and CIDR types. INET\n> is use for hosts, while CIDR for networks. No duplicate code. CIDR\n> behaves the same as INET, except for a different input function. The\n> system catalogs take care of the mapping of CIDR to INET functions.\n> \n> It will require an initdb for beta users moving to 6.4 final. Is this\n> OK? If so, I will apply it tomorrow. If not, the type will have to be\n> disabled for 6.4.\n\nJust to clarify, what we have done (are doing) is to rename the existing\ninet type as cidr. This means that the CIDR type is now for CIDRized\nnetworks as Paul originally proposed. There are functions to extract the\nvarious components such as netmask, broadcast address and mask length.\n\nThe INET type is now for either hosts or hosts plus networks. The\ncode is not quite perfect yet but it compiles and works if you enter\na host as x.x.x.x/32. We'll try to improve it before 6.4 but at\nleast the catalogues are set up so we can fine tune the type without\ndoing an initdb or changing the user API.\n\nBetween 6.4 and 6.5 we'll work on improving the type. While the\ncatalogues won't change, we can modify the underlying code. The\ndecision, which we should really make now for the documentation,\nis what type to make it. Remember that we identified 3 types, INET,\nIHOST and CHOST. With the name change we can call the first one CIDR\nnow. The question is, what type should the new inet type represent,\nIHOST or CHOST?\n\nIHOST is meant to hold a host only. To specify a host and the\nnetwork information using IHOST would require also using CIDR.\n\nCHOST is meant to hold a host and network information in the same\ntype. It can hold an IHOST by itself if desired. There are\nfunctions to extract the various components such as host, netmask,\nbroadcast address and mask length.\n\nAs you have probably guessed, I vote for CHOST.\n\nPaul, can you send me the new functions that you wrote? These will\nbe needed for the new type either way. Hopefully I can fold it in\nbefore 6.4.\n\n> If I don't hear any objections, I will do it some time around noon\n> tomorrow, EST.\n\nOf course, all of the above is assuming that everyone accepts this\nchange.\n\nBruce, I was thinking that since cidr.c consists of a single function\nand it uses most of the code in inet.c anyway, why don't we just fold it\ninto inet.c instead of having a whole file for it?\n\nThis is almost as much work as my day job. And after 6.4 is released\nI get to go through the same thing for the next release of PyGreSQL.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Wed, 21 Oct 1998 07:12:43 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "On Wed, 21 Oct 1998, Bruce Momjian wrote:\n\n> It will require an initdb for beta users moving to 6.4 final. Is this\n> OK? If so, I will apply it tomorrow. If not, the type will have to be\n> disabled for 6.4.\n> \n> The big question is how many people have _live_ data in 6.4 beta\n> databases, who don't want to initdb for 6.4 final, and can't use\n> pg_upgrade to move to 6.4 final.\n\nI do, but since this is my off season it's not a big deal. I usually\nkeep a backup database running on another machine anyway.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n\n",
"msg_date": "Wed, 21 Oct 1998 07:13:13 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> The big question is how many people have _live_ data in 6.4 beta\n> databases, who don't want to initdb for 6.4 final, and can't use\n> pg_upgrade to move to 6.4 final.\n\nHey, sounds like an opportunity for pg_upgrade to get some beta testing.\nI know I haven't even touched it yet (I just initdb'd the last few\ntimes), but I will try it for this if it will work for updating a\n6.4beta database to the new state.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 21 Oct 1998 10:25:55 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types "
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > I have recieved the files from D'Arcy for new INET and CIDR types. INET\n> > is use for hosts, while CIDR for networks. No duplicate code. CIDR\n> > behaves the same as INET, except for a different input function. The\n> > system catalogs take care of the mapping of CIDR to INET functions.\n> > \n> > It will require an initdb for beta users moving to 6.4 final. Is this\n> > OK? If so, I will apply it tomorrow. If not, the type will have to be\n> > disabled for 6.4.\n> \n> Just to clarify, what we have done (are doing) is to rename the existing\n> inet type as cidr. This means that the CIDR type is now for CIDRized\n> networks as Paul originally proposed. There are functions to extract the\n> various components such as netmask, broadcast address and mask length.\n\nThese functions also work for INET too.\n\n> \n> The INET type is now for either hosts or hosts plus networks. The\n> code is not quite perfect yet but it compiles and works if you enter\n> a host as x.x.x.x/32. We'll try to improve it before 6.4 but at\n> least the catalogues are set up so we can fine tune the type without\n> doing an initdb or changing the user API.\n\nWe have to get it working OK in the next day, then any changes go into\npost 6.4 minor releases. We have regression tests and can't be changing\nthings.\n\nI would like the INET type to not display/require the /32 anymore.\n\n> Between 6.4 and 6.5 we'll work on improving the type. While the\n> catalogues won't change, we can modify the underlying code. The\n> decision, which we should really make now for the documentation,\n> is what type to make it. Remember that we identified 3 types, INET,\n> IHOST and CHOST. With the name change we can call the first one CIDR\n> now. The question is, what type should the new inet type represent,\n> IHOST or CHOST?\n> \n> IHOST is meant to hold a host only. To specify a host and the\n> network information using IHOST would require also using CIDR.\n> \n> CHOST is meant to hold a host and network information in the same\n> type. It can hold an IHOST by itself if desired. There are\n> functions to extract the various components such as host, netmask,\n> broadcast address and mask length.\n\nPeople could just put a netmask in the field, so INET seems more\ngeneric.\n\nFunctionality-wise, I like CHOST.\n\n> \n> As you have probably guessed, I vote for CHOST.\n> \n> Paul, can you send me the new functions that you wrote? These will\n> be needed for the new type either way. Hopefully I can fold it in\n> before 6.4.\n> \n> > If I don't hear any objections, I will do it some time around noon\n> > tomorrow, EST.\n> \n> Of course, all of the above is assuming that everyone accepts this\n> change.\n\nNo objections yet.\n\n> \n> Bruce, I was thinking that since cidr.c consists of a single function\n> and it uses most of the code in inet.c anyway, why don't we just fold it\n> into inet.c instead of having a whole file for it?\n\nOK.\n \n> \n> This is almost as much work as my day job. And after 6.4 is released\n> I get to go through the same thing for the next release of PyGreSQL.\n\nJoin the club. It is usually _harder_ than my day job, but it is not as\noften as my day job.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 11:42:18 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > I have recieved the files from D'Arcy for new INET and CIDR types. INET\n> > is use for hosts, while CIDR for networks. No duplicate code. CIDR\n> > behaves the same as INET, except for a different input function. The\n> > system catalogs take care of the mapping of CIDR to INET functions.\n> > \n> > It will require an initdb for beta users moving to 6.4 final. Is this\n> > OK? If so, I will apply it tomorrow. If not, the type will have to be\n> > disabled for 6.4.\n\nINET/CIDR changes applied.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 12:15:08 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> > The INET type is now for either hosts or hosts plus networks. The\n> > code is not quite perfect yet but it compiles and works if you enter\n> > a host as x.x.x.x/32. We'll try to improve it before 6.4 but at\n> > least the catalogues are set up so we can fine tune the type without\n> > doing an initdb or changing the user API.\n> \n> We have to get it working OK in the next day, then any changes go into\n> post 6.4 minor releases. We have regression tests and can't be changing\n> things.\n\nIt does work. More importantly we have the catalogues and API nailed\ndown. There are some effeciencies and fine tuning that could be done\nbut I think what we have is good enough that documentation is now\nmore of a priority than code.\n\n> I would like the INET type to not display/require the /32 anymore.\n\nThat's done. The only thing is that now you have to manually put\nthe /32 on input. I'll fix that very soon.\n\n> > Between 6.4 and 6.5 we'll work on improving the type. While the\n> > catalogues won't change, we can modify the underlying code. The\n> > decision, which we should really make now for the documentation,\n> > is what type to make it. Remember that we identified 3 types, INET,\n> > IHOST and CHOST. With the name change we can call the first one CIDR\n> > now. The question is, what type should the new inet type represent,\n> > IHOST or CHOST?\n> > \n> > IHOST is meant to hold a host only. To specify a host and the\n> > network information using IHOST would require also using CIDR.\n> > \n> > CHOST is meant to hold a host and network information in the same\n> > type. It can hold an IHOST by itself if desired. There are\n> > functions to extract the various components such as host, netmask,\n> > broadcast address and mask length.\n> \n> People could just put a netmask in the field, so INET seems more\n> generic.\n\nThey could. As I have said I wouldn't. If I had to store a netmask\nalone I would store it as a masklen in an integer. To each his own\nthough. What we have gives maximum flexibility I think.\n\n> Functionality-wise, I like CHOST.\n\nNo one has objected so I will go forward on that basis.\n\n> > Bruce, I was thinking that since cidr.c consists of a single function\n> > and it uses most of the code in inet.c anyway, why don't we just fold it\n> > into inet.c instead of having a whole file for it?\n> \n> OK.\n\nWill you do it or shall I?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Wed, 21 Oct 1998 12:33:14 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > > The INET type is now for either hosts or hosts plus networks. The\n> > > code is not quite perfect yet but it compiles and works if you enter\n> > > a host as x.x.x.x/32. We'll try to improve it before 6.4 but at\n> > > least the catalogues are set up so we can fine tune the type without\n> > > doing an initdb or changing the user API.\n> > \n> > We have to get it working OK in the next day, then any changes go into\n> > post 6.4 minor releases. We have regression tests and can't be changing\n> > things.\n> \n> It does work. More importantly we have the catalogues and API nailed\n> down. There are some effeciencies and fine tuning that could be done\n> but I think what we have is good enough that documentation is now\n> more of a priority than code.\n\nOK. Yes, you can continue cleaning things up. I just want some\nregression tests and documentation, and you can then continue, but I\nhope they changes will not affect the contents of the documentation or\nregression tests. If you can concentrate on those changes first, then\nthe docs/regression, that would be good.\n\n\n> > I would like the INET type to not display/require the /32 anymore.\n> \n> That's done. The only thing is that now you have to manually put\n> the /32 on input. I'll fix that very soon.\n\nActually, the /32 is not required. It properly displays it if it was\nsupplied, and supresses it if it is not.\n\nIn fact, it may be supressing the /32 when it should not, like in the\nCIDR type:\n\n\ttest=> create table test (x inet, y cidr);\n\tCREATE\n\ttest=> insert into test values ('1.2.3.4','1.2.3.4');\n\tINSERT 18474 1\n\ttest=> insert into test values ('1.2.3.4/32','1.2.3.4/32');\n\tINSERT 18475 1\n\ttest=> insert into test values ('1.2.3.4/24','1.2.3.4/24');\n\tINSERT 18476 1\n\ttest=> select * from test;\n\tx |y \n\t--------+--------\n\t1.2.3.4 |1.2.3.4 \n\t1.2.3.4 |1.2.3.4 \n\t1.2.3/24|1.2.3/24\n\t(3 rows)\n\nI would think the CIDR type should preserve the /32.\n\n> > > Between 6.4 and 6.5 we'll work on improving the type. While the\n> > > catalogues won't change, we can modify the underlying code. The\n> > > decision, which we should really make now for the documentation,\n> > > is what type to make it. Remember that we identified 3 types, INET,\n> > > IHOST and CHOST. With the name change we can call the first one CIDR\n> > > now. The question is, what type should the new inet type represent,\n> > > IHOST or CHOST?\n> > > \n> > > IHOST is meant to hold a host only. To specify a host and the\n> > > network information using IHOST would require also using CIDR.\n> > > \n> > > CHOST is meant to hold a host and network information in the same\n> > > type. It can hold an IHOST by itself if desired. There are\n> > > functions to extract the various components such as host, netmask,\n> > > broadcast address and mask length.\n> > \n> > People could just put a netmask in the field, so INET seems more\n> > generic.\n> \n> They could. As I have said I wouldn't. If I had to store a netmask\n> alone I would store it as a masklen in an integer. To each his own\n> though. What we have gives maximum flexibility I think.\n\nI can see cases where storing the netmask separately would be nice, and\nit should display in a nice INET format.\n\n> \n> > Functionality-wise, I like CHOST.\n> \n> No one has objected so I will go forward on that basis.\n\nGood.\n\n> \n> > > Bruce, I was thinking that since cidr.c consists of a single function\n> > > and it uses most of the code in inet.c anyway, why don't we just fold it\n> > > into inet.c instead of having a whole file for it?\n> > \n> > OK.\n> \n> Will you do it or shall I?\n\nAlready done.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 13:12:43 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "> I would think the CIDR type should preserve the /32.\n\nMakes no difference. If all four octets are specified and no \"/\" is given,\nthe \"/32\" is assumed. Printing it would be nicer but not printing it is not\nan error.\n",
"msg_date": "Wed, 21 Oct 1998 10:21:18 -0700",
"msg_from": "Paul A Vixie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types "
},
{
"msg_contents": "On Wed, 21 Oct 1998, Bruce Momjian wrote:\n\n> I have recieved the files from D'Arcy for new INET and CIDR types. INET\n> is use for hosts, while CIDR for networks. No duplicate code. CIDR\n> behaves the same as INET, except for a different input function. The\n> system catalogs take care of the mapping of CIDR to INET functions.\n> \n> It will require an initdb for beta users moving to 6.4 final. Is this\n> OK? If so, I will apply it tomorrow. If not, the type will have to be\n> disabled for 6.4.\n> \n> The big question is how many people have _live_ data in 6.4 beta\n> databases, who don't want to initdb for 6.4 final, and can't use\n> pg_upgrade to move to 6.4 final.\n> \n> If I don't hear any objections, I will do it some time around noon\n> tomorrow, EST.\n\nGreat...I'm going to plan for a BETA3 to be put out for Friday, so that\nthe testing broadens out a bit more...\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Wed, 21 Oct 1998 14:43:21 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> > It does work. More importantly we have the catalogues and API nailed\n> > down. There are some effeciencies and fine tuning that could be done\n> > but I think what we have is good enough that documentation is now\n> > more of a priority than code.\n> \n> OK. Yes, you can continue cleaning things up. I just want some\n> regression tests and documentation, and you can then continue, but I\n> hope they changes will not affect the contents of the documentation or\n> regression tests. If you can concentrate on those changes first, then\n> the docs/regression, that would be good.\n\nFor sure not the documentation. As for the regression testing, it should\nbe final by Friday. Unfortunately I'm having problems. The last sup\nseems to have broken the inet type. I'll try to get you in #PostgreSQL\nand talk about this.\n\n> > That's done. The only thing is that now you have to manually put\n> > the /32 on input. I'll fix that very soon.\n> \n> Actually, the /32 is not required. It properly displays it if it was\n> supplied, and supresses it if it is not.\n\nIn certain cases. I now have the functions from Paul so I should be able\nto get that correct tonight if I can get the type working again.\n\n> In fact, it may be supressing the /32 when it should not, like in the\n> CIDR type:\n\nAgain, will be fixed when I plug in Paul's functions.\n\n> I would think the CIDR type should preserve the /32.\n\nInternally it does. Paul said it doesn't matter and I assume it is\nhis function leaving it off though I haven't checked.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Wed, 21 Oct 1998 22:03:53 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > > It does work. More importantly we have the catalogues and API nailed\n> > > down. There are some effeciencies and fine tuning that could be done\n> > > but I think what we have is good enough that documentation is now\n> > > more of a priority than code.\n> > \n> > OK. Yes, you can continue cleaning things up. I just want some\n> > regression tests and documentation, and you can then continue, but I\n> > hope they changes will not affect the contents of the documentation or\n> > regression tests. If you can concentrate on those changes first, then\n> > the docs/regression, that would be good.\n> \n> For sure not the documentation. As for the regression testing, it should\n> be final by Friday. Unfortunately I'm having problems. The last sup\n> seems to have broken the inet type. I'll try to get you in #PostgreSQL\n> and talk about this.\n\nFixed.\n\nThomas will do all the markup. Just some info for the docs are needed.\nHe should provide you with what he needs.\n\n> \n> > > That's done. The only thing is that now you have to manually put\n> > > the /32 on input. I'll fix that very soon.\n> > \n> > Actually, the /32 is not required. It properly displays it if it was\n> > supplied, and supresses it if it is not.\n> \n> In certain cases. I now have the functions from Paul so I should be able\n> to get that correct tonight if I can get the type working again.\n\nGreat.\n\n> \n> > In fact, it may be supressing the /32 when it should not, like in the\n> > CIDR type:\n> \n> Again, will be fixed when I plug in Paul's functions.\n> \n> > I would think the CIDR type should preserve the /32.\n> \n> Internally it does. Paul said it doesn't matter and I assume it is\n> his function leaving it off though I haven't checked.\n\nI recommend displaying the /32 for cidr. Paul says it is not required,\nbut would be 'nice', and I agree.\n\nSorry to be pressuring, but I want this off the Open Items list before\nMarc has a coronary. His head swells badly, and turns all red. :-)\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 22:09:42 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> > For sure not the documentation. As for the regression testing, it should\n> > be final by Friday. Unfortunately I'm having problems. The last sup\n> > seems to have broken the inet type. I'll try to get you in #PostgreSQL\n> > and talk about this.\n> \n> Fixed.\n\nMostly. There seems to still be a problem with the network function as\nI mentioned in the messages to patches. Also, as I mentionwed there,\nthe output of the functions should be left justified, not right. Also,\nI think the netmasklen function should be called masklen. Given the\npolymorphic nature of the type system and the fact that it is being\napplied to inet or cidr types, I don't think there is any confusion\nusing the shorter, more familiar term.\n\n> Thomas will do all the markup. Just some info for the docs are needed.\n> He should provide you with what he needs.\n\nI just received that. I'll get to work. I also have most of the queries\nfor the regression tests which I'll send along shortly.\n\n> > Internally it does. Paul said it doesn't matter and I assume it is\n> > his function leaving it off though I haven't checked.\n> \n> I recommend displaying the /32 for cidr. Paul says it is not required,\n> but would be 'nice', and I agree.\n\ndarcy=> select '1.2.3.4/32'::cidr;\n?column? \n----------\n1.2.3.4/32\n(1 row)\n\ndarcy=> select '1.2.3.4/32'::inet;\n?column?\n--------\n 1.2.3.4\n(1 row)\n\n> Sorry to be pressuring, but I want this off the Open Items list before\n> Marc has a coronary. His head swells badly, and turns all red. :-)\n\nPerhaps we should hand out plastic sheeting like at Gallagher concerts\njust in case. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Thu, 22 Oct 1998 09:57:16 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > > For sure not the documentation. As for the regression testing, it should\n> > > be final by Friday. Unfortunately I'm having problems. The last sup\n> > > seems to have broken the inet type. I'll try to get you in #PostgreSQL\n> > > and talk about this.\n> > \n> > Fixed.\n> \n> Mostly. There seems to still be a problem with the network function as\n> I mentioned in the messages to patches. Also, as I mentionwed there,\n\nI can't find any network anything in inet.c.\n\n> the output of the functions should be left justified, not right. Also,\n> I think the netmasklen function should be called masklen. Given the\n> polymorphic nature of the type system and the fact that it is being\n> applied to inet or cidr types, I don't think there is any confusion\n> using the shorter, more familiar term.\n\n\nDone.\n\n> > Thomas will do all the markup. Just some info for the docs are needed.\n> > He should provide you with what he needs.\n> \n> I just received that. I'll get to work. I also have most of the queries\n> for the regression tests which I'll send along shortly.\n\nGreat.\n\n> > > Internally it does. Paul said it doesn't matter and I assume it is\n> > > his function leaving it off though I haven't checked.\n> > \n> > I recommend displaying the /32 for cidr. Paul says it is not required,\n> > but would be 'nice', and I agree.\n> \n> darcy=> select '1.2.3.4/32'::cidr;\n> ?column? \n> ----------\n> 1.2.3.4/32\n> (1 row)\n> \n> darcy=> select '1.2.3.4/32'::inet;\n> ?column?\n> --------\n> 1.2.3.4\n> (1 row)\n\nNice.\n\n> \n> > Sorry to be pressuring, but I want this off the Open Items list before\n> > Marc has a coronary. His head swells badly, and turns all red. :-)\n> \n> Perhaps we should hand out plastic sheeting like at Gallagher concerts\n> just in case. :-)\n\nYep.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 10:17:45 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "I've added in the inet/cidr types to the regression test, using mostly\nthe tests that D'Arcy sent. Just changed the layout a bit to conform to\nthe other tests and added a test or two to do equality comparisons.\n\nD'Arcy et al, could you follow up and augment those tests, either now if\nit's easy or for v6.5? Probably better to wait a bit as we learn more\nabout the types and their expected behaviors.\n\nAll regression tests pass on my Linux/i686 machine.\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 19:58:58 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "Thus spake Thomas G. Lockhart\n> D'Arcy et al, could you follow up and augment those tests, either now if\n> it's easy or for v6.5? Probably better to wait a bit as we learn more\n> about the types and their expected behaviors.\n\nWhat kind of things are you looking for? I thought I had most possibilities\ncovered in the tests I sent.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Wed, 28 Oct 1998 05:49:20 -0500 (EST)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "> What kind of things are you looking for? I thought I had most \n> possibilities covered in the tests I sent.\n\nThe original test loaded a table and displayed the results. I'd like to\nsee some queries which exercise the operators associated with the types.\nIn this case, that may just be comparison operators, but if you have\nmath operators we should exercise those too. And any other conversion\noperations, such as converting back and forth from int4, if that is\nallowed.\n\nSo, things like (don't know if this exact example is sensible):\n\n select '' as count, c as cidr from inet_tbl\n where c < '128.149.50' and c > '128.149';\n\nwould be interesting.\n\n - Tom\n",
"msg_date": "Wed, 28 Oct 1998 15:19:30 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
},
{
"msg_contents": "Thus spake Thomas G. Lockhart\n> So, things like (don't know if this exact example is sensible):\n> \n> select '' as count, c as cidr from inet_tbl\n> where c < '128.149.50' and c > '128.149';\n\nOf course. I only sent in tests for the functions I added. Ok, add\nthe following two INSERT statements.\n\nINSERT INTO INET_TBL (c, i) VALUES ('10', '11.1.2.3/8');\nINSERT INTO INET_TBL (c, i) VALUES ('10', '9.1.2.3/8'); \n\nThen add the following query.\n\nSELECT i, c,\n i < c AS lt, i <= c AS le, i = c AS eq, \n i >= c AS ge, i > c AS gt, i <> c AS ne,\n i << c AS sb, i <<= c AS sbe, i >> c AS sup, i >>= c AS spe\n FROM INET_TBL;\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Thu, 29 Oct 1998 08:53:05 -0500 (EST)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New INET and CIDR types"
}
] |
[
{
"msg_contents": "Just a clarification. I am not changing the system columns, just adding\nrows to certain system tables. If you don't use the INET/CIDR type, and\ndon't want to run the regression test, you should be fine, even if you\ndon't initdb.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 02:20:47 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "New INET/CIDR type"
}
] |
[
{
"msg_contents": "Hope this is the last one,\n\n Bruce, please apply.\n\n The patch does 2 things:\n\n Fixes a bug in the rule system that caused a crashing\n backend when a join-view with calculated column is used\n in subselect.\n\n Modifies EXPLAIN to explain rewritten queries instead of\n the plain SeqScan on a view. Rules can produce very deep\n nested plans, so I lowered the indentation a little.\n\n The patch is regression tested. Now the rule system handles\n this:\n\n DELETE FROM shoelace WHERE EXISTS\n (SELECT * FROM shoelace_candelete\n WHERE sl_name = shoelace.sl_name);\n\n Doesn't look that complicated. But:\n\n View shoelace is a join of 2 tables with a calculated column.\n\n View shoelace_candelete is a qualified select from\n shoelace_obsolete.\n\n View shoelace_obsolete is a select from shoelace with a\n subselect qualification from shoe. The subselect has a\n qualification that compares calculated columns from shoe and\n shoelace.\n\n View shoe is a join of 2 tables with 2 calculated columns.\n\n\n AND IT WORKS! The generated plan is:\n\n Nested Loop\n -> Merge Join\n -> Seq Scan\n -> Sort\n -> Seq Scan on shoelace_data\n -> Seq Scan\n -> Sort\n -> Seq Scan on shoelace_data s\n SubPlan\n -> Merge Join\n -> Seq Scan\n -> Sort\n -> Seq Scan on shoelace_data s\n -> Seq Scan\n -> Sort\n -> Seq Scan on unit u\n SubPlan\n -> Merge Join\n -> Seq Scan\n -> Sort\n -> Seq Scan on shoe_data sh\n -> Seq Scan\n -> Sort\n -> Seq Scan on unit un\n -> Seq Scan on unit u\n\n\n\n\n What a plan. 6 scans over nested subselects for one DELETE.\n I think that is close to the rule system we wanted.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\nbegin 644 rule_fix.diff.gz\nM'XL(\"(7)+38\"`W)U;&5?9FEX+F1I9F8`Y1UK<]LV\\K/Z*Z#<M-6#BMZ.93>>\nMR:69.]^D2>LX;;]I:`J2>:%)'4G9]67\\WV\\7`$&0!$2*EIQD+M/*(@GL&\\#N\nM<@$MW.62])R01*'S/`C=5?_*=CY1?]%W@IL;VU]$??K7VK-=_[F#;;8\\_J[3\nMZ50!T_B#+LA[)R:C(1D.3J;#D_&(#&>SX^]ZO5X)CDS?\\<GXQ<E@P/MVLO\\8\nM,:,C:SPD[!)!X^6(P+?>=X3\\S?4=;[.@Y*=@';LW[G]IV`<DOD_#Y]=GF0:V\nMX]`HZO]E.W'^T29VO:@?4L^QG6N*3[O*TY#>A6Y,^^+O/X$7+P$/_\\7W:[J@\nM2Q+%X098>L.9_!#;,86GG[4LC5]8DU'*TOC8FDP3EAXR($XYD@B^NPYQKNV0\nM=,3S^:_`Z&7P(0Y=?]7\"\"])!WJT,`-*A4?L4&!(@;@-WD31X[]/?-C2\\;[%/\nMTOD/_K'(51!XY):&5T%$+?*:J^]G&L5D`1\\,&*>JWX$/TDG`<2@]+<>3(VMZ\nMQ#DFC`;XHW;;C00FV\"9IH#P:2$)$7I)W']^^M?!N@]\\:G>(%RH5=HVC8G;QT\nMV$W7CZ&?1WTA\\(:[)*WSZ-55$,9T<1G:?@2&XP;^W[W`^<3ZMMI(1^,S4^%D\nMAOP)%>Z-O[=N%#<ZW/!BRLGG][P:=.KT<C2QCF:I)<+E[#AAH]$(:;P)F4@:\nM#XBO2QK]#KG@`X'$UV&P65V3<.-1$MU',;TAG3XVDA2#6ACCHDN+L<\\MB,,Z\nM]P$,)8X=41(L\"=CP^;L/EV]>_4S>O;_\\Y_F[?U@DIIY'[)AXU`;YQ-?PE6-!\nMUA5,8`'G;]OXX#-^-*@7K%H`Y?SU&XL\\8YB)&,,1B0/B!_$UC)QGC)R4UR[R\nM*ND3.B3T%KNOV2!CR)=!2&&V:'D6D30PY(W\\X/*6;AC%+:]MI1J7`TF@@I'4\nM5492TA<&$]X&^UW#((])L(F9N.A?U-F@ECE%0`L)?$H8CZP'^^AG1SU</6K<\nM0W\\F6,V@ZZJ#KEL8=%W=H.MF!AV;4E#@C\"'D$8ER05.M/__\\D]S1'V\\I<5<^\nM2'V1:+'9;\".3T(]U>DG$S\"^M#\"Q6NZ`,1M9H,$MF(R#A,__3N+#]%;V\\\\M[X\nM,<H%AA1`]>/K5JOUP4'5MQF.=N\\L@DM8,&`N[9&A16C4.PMC^\\JC'\"V;A1H1\nMT]JR=;59@@6\"BLCWT3.PEYA\":[KT[1MLSMO:ZS4LEGPR/_>700O6$U#)9LD`\nM0H,'_N<*S.X3OP6KCKWQXA.^+(Z'UF@R3L?NH9C2$\\JX>R:YP:$)]YV;=4OE\nM5O+N,=Y)\\R49M$67S^)O3FK?1^19KEN\"9*O,>).'K43G%%$F9JTMC<&6QL>*\nM+>%@;+D@XL$I<<E/Q/47U(_A>[?;%AHQB9`)D#4QMB\"D=T9200N7:@X3P[M@\nM07DST.OFBOL#;9+,/C\"`0</</^`DD2Z9H)(%3LYXJG/@(+)OZ6+.+\\4BP#VP\nM(VOT8J28VK?$\\VAGGA?EWG5('7=-3<YU\\K3$MTZ:U7\"MM5V'X%:?#$9FSWH\\\nM'5CCZ0O%>&$)\\BF`@.4Q6\"XCRI>;6_`Z?1!U!%(4<[,-2[(/[>QP16,/7!+9\nM_XHZ]B9\"[X#>$SND3/)_`4P`R#HC\\ZX/`+B0DXY]KMWW#.WO=HBZC5HM_(,Z\nM=?WU)OZM=X:H<>3.8=E8Q=>)491UXX2B[Y3MS)VH!JX\\W/X8B2%.F821%Y%E\nM&-QP)^7:]18AN!I,%@V^](#CD)B.[P2^T^*7%L<\\E_,G\\_.%O-.!\\\\W)VX()\nM^S$BY_V?3.H5ANZM2^],`Y<_*QFVO%&-0:OIR(;L:+HE&![.K-$H76]8\"(9F\nMP*:\\Q(AL\\)A3([JZ)R/>BNG<I#<DYU<[C&A6=2.N[UPOI3&WD)'T$<PMK^U;\nMF-M_4]L3X??QA44PEXR/)V!.VG,I?])RJ[.H&+M@QVR/ABQ#UC!-C?06:FJM\nM,=69WE0K01B?#,8G,*L9;79R;$VGRB(#D0$+-3JN#S&CO9@O/7O%!\"6B%?:4\nMS0:7.(Z9XSJ'.6XAU(F:QZ$>X[S\"ICM^%6V@]2WUHOEF78!GQ^!97&UB6@42\nM-/8#+=!>)I'\"37(.YLBLL0\"5/Z\\`\"((J8-<,*$N>3^^J\\,T@WP1@=/>O5JN0\nMKCZN63L%^M;FKQE1OV=I8MU^19]K?W3]'`9K'#,%%/P:K9J+`L+($,)'^!1+\nM*MJ6\"'ITIC<<O+\"&0\\7X7'`LT=N#A8.O/*WD#I+,OFT\\.N<YE&SXDW9'-IFF\nM`,JO%Q]&\\W=O_IC__NKBW7OP+^52QY?)W%2AP:'W:,RK:P&`R;?9CKZ(%AF4\nM*B]'RP\"($8W!/I/$ZX\\7%V_>71:E82FF,M@=F<(CT_MP,`7%*N'NUZ_8U'5Z\nMK&Z-D,SJK27T+Z9AW5\">3H?6;#A)AK+,0/<P;Y:;ATF/W89EBKO7-N;Z:(C)\nMK0BS9L&2Y\"37)G?7X&.NPP#?$L`2S@%<G%^P]&;$+OO921NN:L__T/<S_,\\2\nM)=B<I2\\_OGV+#])<9(_@973GQLXU:W9IK]C?=IMW;[#$Z>7\\D@F/+90G['Z#\nM/VXTE\">-3LP\\YI9R#Y2!\\!)4\\\"_/DKC=:\"3*:P&4WAF;@RWY4#`JKPN+'4NM\nM])34BD`H&.#K0(YV?K/1L5<K))M?UJ$8`\"0&MD>:_Q$&F_5K#Z.O'.'*DT9G\nM%:Z1>N5>'18`\"@@=5;9'#G`YS9&.MQJXPB+-?+G=G5CHWCL#<4=[I/4\\IGE:\nM\\5:C`XYIB,3B51UBL7_O##_W;=-A:-]?T&7>JL5M?+&S9'8M;M0A'D\"P%.8&\nMUKR03;X[L[`#%B^X>P(LAX5N1]'*W[.F81G)*1GN-#H8I8)^?[<U=HD3/SSO\nMG<$'Q[59XR*004T2<(VD*;@FW9=B63DM/@R\\1?'Y0SD#$+W:-R=;&KP.(%;;\nMU@`7[IP,Y)M*R;1X:4;PK5D+'PNYM!-.C&J4>DS?IJ7/\\NK;67\\?-E=O7?]3\nMCGYQM]$!<*A'<5UGF`*$WIE'ES'X0XM=#&\\W#/`148\\Z55<Y<-^&5>3#WA/F\nMI,/N-3K_`1\\\"9\"/>).XN&>BO^GX'D0W#@<[LX:\"G>9_#X5BE'L0^YJ[T]2%[\nMJY)Y6[[Q/_G!'<^YDMA>D>\\7F(O.$]I^9I&L6WJJ@\\:X<*,3_OJ3=4C*5C*]\nM5/KP#I+](\"\\37S^7*DE\\?4R/)WY^+NQHDS@(#![\\_A(O7Z]+G^>QFDLO.9=W\nM4NZ_B)]?SH;>SS\\((WMT_LOYTCK_!V&K7D10SH$N(C@(`S7#A'(.C&'\"8<;)\nMHV.'<HY*8H?:?.V`VA!0/`7J+X!2%WH<Q'H.%(^0'WZ0)\"D!\";1*N-!'+\"_3\nM==,0LVA:?$-1B]$\"MD8M1<WK5/_4L4RY-9MBF7J6O!M:78\"S,]Z#1SWES)BC\nMG@-+41<*/05*;7ST%(CU0=-^9]T:D52>^J>.I%C=ON;5,.FQ1PU1EPU2\\>Z)\nM`]!BNF!K`4X:6.YG^\\1F';&@UPF\\S8U/\\,4Z,,<A)'5+[M)U;'S%8Y$[2FXV\nM,'W:BW]O(EGJ)%<;$`O6=L!M]HIOBC7UT\\F6E[M'TZDU/#H:I/7KP]$,;DV&\nMZ6X*\\:HHN_\\!+NWUVKN?7YQ?S#DU\\T3;A0@P&^2)(NABD)>M(<=+4Y#7-01Y\nM73[1='<*\\C@J7CYI8$@T,(1[7;VU=[FU=Q5K$JARD5R&ZNV17\"U:LS%=76HS\nMX5J&Y*KA6BWB,X%;7=IY3)8AVAR3U2)3C<[J4LD#KPR56P*O6F060K#:MBNC\nMJZSUED97M<@VQ%E&XA^!(A=/'0+%`4&K\\5%=U;+0)Z-54^@C22V&/9DYOP)2\nM$8B8&XA`Q-R`!R(9NF4@(@DU!B(DZ5-%[+J@I&N(-KHB%#-+0/H^W<J^3ZX2\nM3^?Z=&NY/@7RDLU<8I=5M@(-KE)I80EG28%;<1^-1;@26$$V;]YA/I5+%SK?\nMP;0==/AB-K#@0]V2Q#<G)-_Y3.\\$Z_OW5_^&\\`=G['0?C#9V/R,#C->;,K84\nM(L:>Z+6PH=(F<MM-HR5&!S;(CX<\\=(GZ09*8B`U:`X13>3L1\"-R_O/CX)GG`\nMMY(,IQ/PUZ93=3/)/CAO2Z[,`P(`607&VJ>U^=%J]O@8^#N>C13-5ALC6<O4\nMC1%2:XR00G@@-G<RH3.='(V/P:V>;\"NKG&&3V20UV#66/\\<AI<I6FG03#;;0\nME-\\I;4!=^9=+LEA,J6<CFB*X[?TX1MT.!+Y_.!.%Z5#*FN4=R]]*P0O*ZL)/\nMM_^&U&,!UIP9<5LH\\7@&4\\I@(`?6'C1DJELL4=#6;EOU8RH?W(]ZMD(_F':T\nMNSI&DR-K-)HJTX2$#\"K@7UI*S:JR_T54NR8;8'`I>@VS94M7[BKZ8J[#8`F:\nMHE)L7:E8.&VX<Z'H@]C]BWM]P7KOKL'%)\\XF#''/WB*@D?]C3.Z\"D.<>Q+Z!\nM9`?W<_A'[BG;%LYL?S9[88T&`Z4Z^*N6IK%<]FL0J'[3Z_$$=[TJ!P>,AN.I\nM-1H>R86<#QPV`%31B<U/E_=KECIY_<O/<['1/]DHZ@1^[/H;Q4'O=WH[_Q-;\nMTCOD#\\J<!U)(1T58!8QWY<0H^\\0!>\\!I9OON;EV'LNIA^SGY$*#,G+3YM1V)\nM='I$@&'<B<]!W^`CXL;/DZ:RRRN&_>CY!%I[]RD)>1(=VU?0W-(D!1<])^>(\nMPHU$)C%*LVLW]B>%%6QB<^_\"\"X)/X![9X@P!'[>OB+WTJ/0\"D?]*D==70)]_\nM04-05B!59#_\\0)H9&U$>MK,AD=(LZQ0J#V348H(I'3<UJ$'ZFF8\"*])G`)!#\nMF.Y^K\"'69-OGAYBNR?\"D^HZR7VS?76_?3R::5-I-)MK6WDNFZP^=QR?#+4<!\nMX69[=9&$95F)V0R;CK@OD1S+TLQ7_NM+]GFP1G2)7O5@E&:2Y26Y-\"\\^X0<P\nMZ#*]XGR&9C[32_297L(3%O((AAP',IMK)<0;SS+(Y6XS=(C<+2;O\\^G;$@*4\nM%&T9\"?SU!\"OEXG9\\B1,43I,PZ5Y16)I@]H2O.(G)(T`VN\"&#L#<Y#DO*1FSI\nMDC!@-L`H*4IN]%5>,YG?#,.9+\"^#7LS]EG`N?<86ZR]SO545P5.[&:I8:A>5\nM@!I5TKM5*<%N/)E;F0J>NLU0P5*W2$4N>UN5BFRJMKIARL1LUC23Q\"Q25,S-\nM5B5*EXA52*O664VQ[MRY5B>9%JTL1I8$S4@0DZ!,>IE$:%YP^BKS_+-<D?EV\nM4GAB,T,+2VPB+9Z\"6CE/*4EM$EYD(3(Y.2DIIRI5%(HLC,@0DQ1\"L-.+Q/O-\nM3'U$7D8P=_$OW+FTO2C@KAG,6O;\"7LOWFBX_R\"!Y*2H[\"2QI#07S,.7=`&Q3\nM-NZ733L%8.V,>6EH_@5GUF@3LA,CQ,YZF\"TV-Y0M?EE\"R#J`-1$GYQ0`=HOL\nM&WY,!4S<17XPOB\"L$O?.!='?T;3SG>MYX$WYN-2\"EPV3IA`2SO#RA7)$<$T!\nM*V3-(2I)`<`:@8<`.`&$,D[LW;<+LA*6T9(3I[B180P/AF$OO,C+Q,#RS:1$\nMRT\\&:HK#+/@!5?S4#60_2M8BI==#^E9:.\"2YJ*_@D'@+GD)*2Y\"5RV*&^2MU\nM6G)<JDY+RJ#\"7,&/.Z1#DR,NZ]#4)N];=W9,\";*\"LZ.34/KF`11025K[\\8M,\nM1&?\\HL=;W\"-\\)N.FZ+S/M(=QL1=_RI@HU?M35<BN!CCG:^T/\\-X!JC[:X]56\nMUW]37X?QPEQ)C/(NL'3;8?Y$OD(I+_M^6GR<EO**%D0MY=WRMJZG]?'2TWML\nMD.N85_\\]D;N94[CJ;CY>OX]S19M*)2JVY4>;*N6HJ4LFZV4KC&8&)CFU8K<I\nM7534?ED?V;A::7WD79>L_P=?.B-2#;LU/>NTI\\ZW3G$>TLLFY-5BP8X8$H-$\nMIFJ30X98430_IF]DC6=*1<(.648F+>!73F3)%L4FX43NDH-\\A(.?^OC%\\M-=\nM_?NMY:>FG$M:5:*K.6WF]JXV]<5>&H/HDJINOZ$&M3K!V<+3^B1G/=M*'OK6\nM0M3J'&2J3_<E<Y-KKJU&K4ZK6H*Z+U)-/KJ^)+4ZK84ZU+W9<XF[;JQ+K4Z[\nMH1BU(@<[(LF5HQX&R4&!JR6I^]*RWKO7E:AJ'?MMIX4TMYT6TBPY+20Y,EO/\nMP./+72NXZ3H?7;/O3I\\23MTW385K,W]82--0\\EJJOJW.NW/R\",WOKZML\nMV#LA4'?25<$@7/PF24]5%](1B+.;YIJJ;+9MFJM.=G'/W-XEH^Z..PSPS#ZX\nMRA/*3BBR.]X>/6EEW>]*Q:OYL_4TQ:O-1Q>O)A7=J:N?^26-;#0H?DBC7SGI\nMOL.Q'U^M2YYCL9I+GNZ,;!9W1GX)/[V4\"[V?OB<^#N:\\E[*E==X/PE4]C[Z4\nM`9U'?Q#KJNGFES)@=/,/,T8>[?N7,E3B^]=A:T?,AH#@\"3`_/49=Z'`0RSE0\nM/*'N'C*?]M'<?MI'L_2T#_*M11TF`]@:=105K]/\\$\\4BU6<,4RQ2;_CLA%47\nMH.R,UARUD/U$+:6\\F*.6P\\I0%\\H\\`49M?/,$>/5!SWZGVQJ14/X@PJ>(A)++\nM75]]%\"O)Q4^@ZDNWKUEU]_8FV4KR[6V+E>33HVPE^6[]QR>C+;]),CRV1O(L\nM>?E;K\\],O_7Z+#EN/OGQRZ2:GKTJ0S$7HG-#7?FI4$OMJB^Q`X#H7T-Q4BHI\n<G/^`!A=$_N=T#R\\(RTS_UR\":_P%H_<@64GD`````\n`\nend\n\n",
"msg_date": "Wed, 21 Oct 1998 14:18:50 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "One more bugfix"
},
{
"msg_contents": "Applied.\n\n\n> Hope this is the last one,\n> \n> Bruce, please apply.\n> \n> The patch does 2 things:\n> \n> Fixes a bug in the rule system that caused a crashing\n> backend when a join-view with calculated column is used\n> in subselect.\n> \n> Modifies EXPLAIN to explain rewritten queries instead of\n> the plain SeqScan on a view. Rules can produce very deep\n> nested plans, so I lowered the indentation a little.\n> \n> The patch is regression tested. Now the rule system handles\n> this:\n> \n> DELETE FROM shoelace WHERE EXISTS\n> (SELECT * FROM shoelace_candelete\n> WHERE sl_name = shoelace.sl_name);\n> \n> Doesn't look that complicated. But:\n> \n> View shoelace is a join of 2 tables with a calculated column.\n> \n> View shoelace_candelete is a qualified select from\n> shoelace_obsolete.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 12:21:16 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] One more bugfix"
}
] |
[
{
"msg_contents": "Hi,\n\n while writing the chapter about Rules and permissions I\n remember that there was a problem with non privileged users.\n As soon as someone without superuser privs does a GRANT or\n REVOKE on his relations, he must GRANT explicitly to himself\n too or will get a \"permission denied\". I think since the\n table owner allway has the right to change ACL's, this\n doesn't make sense. I'll dig it up and send in a patch soon.\n\n While doing this, should I exclude RULE permission from GRANT\n ALL? I think it's dangerous to have it included, because the\n usual way to give full access is a GRANT ALL and someone\n might forget that this includes the right to disable rule\n actions for a moment. The output of pg_rules gives anyone the\n knowledge to reinstall the correct rules after. An explicitly\n required GRANT RULE is better IMHO. And the RULE right isn't\n standard, is it?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Wed, 21 Oct 1998 18:13:50 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "ACL's"
},
{
"msg_contents": "I think it should stay that way - being able to deny oneself a privilege is a\ngood way to make sure that one does what one does consciously. I know the\nroot password on many machines, but I still do almost everything through a\nnormal account - that way I have to make a conscious decison to become\ndangerous :-) and if I accidentaly try to do something dangerous as an\nordinary user a) it doesn't happen and b) I'm reminded how dangerous it is.\nI still have the ability to do dangerous things, I just have to take an extra\nstep.\n\nI agree with your point regarding RULE permission and GRANT ALL; however,\nGRANT ALL really should grant ALL, don't you think? Maybe add a variant\n\"GRANT NORMAL\", where \"NORMAL\" is a mask of permissions set by the\nadministrator (of the given database of course).\u001b\n\nRegards, K.\n\nAm 21-Oct-98 schrieb Jan Wieck:\n> Hi,\n> \n> while writing the chapter about Rules and permissions I\n> remember that there was a problem with non privileged users.\n> As soon as someone without superuser privs does a GRANT or\n> REVOKE on his relations, he must GRANT explicitly to himself\n> too or will get a \"permission denied\". I think since the\n> table owner allway has the right to change ACL's, this\n> doesn't make sense. I'll dig it up and send in a patch soon.\n> \n> While doing this, should I exclude RULE permission from GRANT\n> ALL? I think it's dangerous to have it included, because the\n> usual way to give full access is a GRANT ALL and someone\n> might forget that this includes the right to disable rule\n> actions for a moment. The output of pg_rules gives anyone the\n> knowledge to reinstall the correct rules after. An explicitly\n> required GRANT RULE is better IMHO. And the RULE right isn't\n> standard, is it?\n> \n> \n> Jan\n> \n> --\n> \n>#======================================================================#\n># It's easier to get forgiveness for being wrong than for being right. #\n># Let's break this rule - forgive me. #\n>#======================================== [email protected] (Jan Wieck) #\n> \n> \n> \n\n---\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nKarl Auer ([email protected]) Geschaeft/work +41-1-6327531\nKommunikation, ETHZ RZ Privat/home +41-1-4517941\nClausiusstrasse 59 Fax +41-1-6321225\nCH-8092 ZUERICH Switzerland\n",
"msg_date": "Wed, 21 Oct 1998 23:12:53 +0200 (MET DST)",
"msg_from": "Karl Auer <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] ACL's"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> I think it should stay that way - being able to deny oneself a privilege is a\n> good way to make sure that one does what one does consciously. I know the\n> root password on many machines, but I still do almost everything through a\n> normal account - that way I have to make a conscious decison to become\n> dangerous :-) and if I accidentaly try to do something dangerous as an\n> ordinary user a) it doesn't happen and b) I'm reminded how dangerous it is.\n> I still have the ability to do dangerous things, I just have to take an extra\n> step.\n\nWhat do other DB's do. I assume they give the owner permission.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 17:36:17 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] ACL's"
},
{
"msg_contents": "> \n> [Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > I think it should stay that way - being able to deny oneself a privilege is a\n> > good way to make sure that one does what one does consciously. I know the\n> > root password on many machines, but I still do almost everything through a\n> > normal account - that way I have to make a conscious decison to become\n> > dangerous :-) and if I accidentaly try to do something dangerous as an\n> > ordinary user a) it doesn't happen and b) I'm reminded how dangerous it is.\n> > I still have the ability to do dangerous things, I just have to take an extra\n> > step.\n> \n> What do other DB's do. I assume they give the owner permission.\n\n Hmmm... so it's a TODO for 6.5 after beeing discussed.\n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n",
"msg_date": "Thu, 22 Oct 1998 10:50:17 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] ACL's"
}
] |
[
{
"msg_contents": "> It's been quite some time since it was necessary to run the postmaster\n> with a specific TZ setting in order to do the regression tests. I see\n> that the instructions to change TZ have been removed from install.sgml\n> (though not yet from the text INSTALL file). But the instructions\n> still\n> tell you to start a postmaster for the regression test, then kill it\n> and\n> start another one for production use. Seems to me we could just\n> rearrange the order of the steps: start the postmaster normally,\n> *then*\n> run the regression tests if you feel like it. Is there a reason not\n> to\n> simplify the instructions that way?\n> \n> \t\t\tregards, tom lane\n> \n> PS: Actually, what we really need is an easy way to run the regression\n> tests on a new build without having to disturb the production\n> installation until you know the new build works. (Everyone else does\n> \"make test\" *before* \"make install\" ... with pgsql you have to do it\n> the\n> other way around. Not good for mission-critical software.)\n> \n> Right now I think this requires configuring, building, installing with\n> nonstandard values of POSTGRESDIR and PGPORT, then running the\n> regression test in that environment, then (if successful) throwing\n> away\n> all that work and repeating the build with standard configuration so\n> you\n> can install it live. Bleah. And you still run the risk of blowing it\n> by misconfiguring the second time around.\n> \n> What I really want to be able to run the regression test on the\n> software\n> sitting in the build tree, without doing an install at all. Any\n> thoughts on what it would take to do that?\nWhy not just use the -D and -p postmaster options to the current\nconfiguration to supply a test port and a DataDir in the src tree. The\nDatadir would have to already be initdb'd (I'm not sure how this would\nwork).\n\t-DEJ\n",
"msg_date": "Wed, 21 Oct 1998 11:32:42 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Anyone object to simplifying INSTALL instructions?"
}
] |
[
{
"msg_contents": "\nOn Friday morning, I'm going to be tagging the PostgreSQL source tree as\nSTABLE-6_4, and will be creating a new BETA tar ball. \n\nNormal development will be able to continue, while we will be able to\nmaintain/fix any bugs from those testing the BETA.\n\nOn November 1st, I will be tagging it once more as RELEASE-6_4, which will\nallow us to cover over any bugs as reported by those upgrading to v6.4.\n\nStarting Friday, if anyone wants to start adding in any new features or\nwhatnot, this will be possible...AFTER I post that the tagging is done!!\n:)\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Wed, 21 Oct 1998 15:42:28 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "CVS Branch Tagging..."
},
{
"msg_contents": "> \n> On Friday morning, I'm going to be tagging the PostgreSQL source tree as\n> STABLE-6_4, and will be creating a new BETA tar ball. \n> \n> Normal development will be able to continue, while we will be able to\n> maintain/fix any bugs from those testing the BETA.\n> \n> On November 1st, I will be tagging it once more as RELEASE-6_4, which will\n> allow us to cover over any bugs as reported by those upgrading to v6.4.\n> \n> Starting Friday, if anyone wants to start adding in any new features or\n> whatnot, this will be possible...AFTER I post that the tagging is done!!\n> :)\n\nWhile it will be possible, I would like people to concentrate on the\nexisting 6.4 issues we have, and discourage parallel development on\nseparate final and features source trees. If 6.4 is similar to previous\nreleases, we will need several minor releases before we can say goodbye\nto 6.4, and start working on new features. Of course, certain minor\nfeatures can be put in minor releases.\n\nCerainly a Linux person can get set_proctitle working for the ps args. \nMassimo has done most of the work in breaking out that functionality.\n\n---------------------------------------------------------------------------\n\nAdditions\n---------\nnew functoins and test INET address type(Tom Helbekkmo, D'Arcy)\nCREATE TABLE test (x text, s serial) fails if no database creation permission\nregression test all platforms\n\nSerious Items\n------------\nchange pg args for platforms that don't support argv changes\n\t(setproctitle()?, sendmail hack?)\n\nDocs\n----\ngenerate html/postscript documentation\nmake sure all changes are documented properly\n\nMinor items\n-----------\ncnf-ify still can exhaust memory, make SET KSQO more generic\npermissions on indexes: what do they do? should it be prevented?\nallow multiple generic operators in expressions without the use of parentheses\ndocument/trigger/rule so changes to pg_shadow create pg_pwd\nlarge objects orphanage\nimprove group handling\nimprove PRIMARY KEY handling\ngenerate postmaster pid file and remove flock/fcntl lock code\nadd ability to specifiy location of lock/socket files\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 21 Oct 1998 16:25:43 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "> While it will be possible, I would like people to concentrate on the\n> existing 6.4 issues we have, and discourage parallel development on\n> separate final and features source trees.\n\nAs Bruce describes, a \"quiet time\" just before and up to a month after a\nmajor release is a _really_ good idea. For a couple of reasons:\n\n1) it lets us put out bugfix releases with a minimum of trouble\n\n2) it lets people planning larger changes for the next release to work\non a stable, quiet tree, with some assurance that they will likely be\nable to remerge their work when things unfreeze.\n\nimho we made a few too many changes (leading to uncertainties in exactly\nhow much had changed, and uncertainty in how much might be newly broken)\nin the v6.3->v6.3.1->v6.3.2 release series, and we should try to be even\nquieter this time around.\n\nBruce, can you add the following items to the ToDo for docs:\n\n> Docs\n> ----\n> generate html/postscript documentation\n(User's Guide, Administrator's Guide, Programmer's Guide) (Thomas)\n> make sure all changes are documented properly (All)\n\nmarkup and merge JDBC docs from Peter (Thomas, others??)\nmerge the release notes into doc/src/sgml/release.sgml (Bruce, Thomas)\ngenerate new text-format INSTALL and README from sgml sources (Thomas)\n\nThanks.\n\n - Tom\n",
"msg_date": "Thu, 22 Oct 1998 05:24:50 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "> > While it will be possible, I would like people to concentrate on the\n> > existing 6.4 issues we have, and discourage parallel development on\n> > separate final and features source trees.\n> \n> As Bruce describes, a \"quiet time\" just before and up to a month after a\n> major release is a _really_ good idea. For a couple of reasons:\n> \n> 1) it lets us put out bugfix releases with a minimum of trouble\n> \n> 2) it lets people planning larger changes for the next release to work\n> on a stable, quiet tree, with some assurance that they will likely be\n> able to remerge their work when things unfreeze.\n> \n> imho we made a few too many changes (leading to uncertainties in exactly\n> how much had changed, and uncertainty in how much might be newly broken)\n> in the v6.3->v6.3.1->v6.3.2 release series, and we should try to be even\n> quieter this time around.\n\nYes.\n\n> \n> Bruce, can you add the following items to the ToDo for docs:\n> \n> > Docs\n> > ----\n> > generate html/postscript documentation\n> (User's Guide, Administrator's Guide, Programmer's Guide) (Thomas)\n> > make sure all changes are documented properly (All)\n> \n> markup and merge JDBC docs from Peter (Thomas, others??)\n> merge the release notes into doc/src/sgml/release.sgml (Bruce, Thomas)\n> generate new text-format INSTALL and README from sgml sources (Thomas)\n\nAdded. I know I need to give you a new HISTORY file, and I will do that\nthis weekend, hopefully.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 01:32:28 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "Thomas, can you check the pg_proc table, and remove any of the functions\nthat are no longer needed. Seems we have a lot of them in there that\nused to do conversions, that are no longer needed.\n\nThere are even functions to convert from int2 to int2. We are also\nrunning low on systgem oids.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 01:56:19 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "I see it now. There is parse_coerce.h. I need to use that for the CIDR\ntype, because it is the same as INET inside.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 02:06:46 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "> Thomas, can you check the pg_proc table, and remove any of the \n> functions that are no longer needed. Seems we have a lot of them in \n> there that used to do conversions, that are no longer needed.\n> There are even functions to convert from int2 to int2. We are also\n> running low on systgem oids.\n\nIs this a problem for v6.4? Perhaps we can go with what we have for now,\nand do a cleanup early in v6.5 development?\n\nOne thing: I've been thinking that, although we now do \"automatic\ncoersions\" of types which will try to do the right thing by converting\nexpression or function arguments to the appropriate type if there is not\ndirect match in the catalog, a direct match is always/usually better to\nhave. So, for example, there is an operator defined to multiply an int2\nby an int4. If one actually had an expression that was\n\n '2'::int2 * '4'::int4\n\nthen this operator would be used. If one removed this operator, then the\nexpression would still be evaluated, but done as\n\n int4('2'::int2) * '4'::int4\n\nwhich in this case will give the correct result, though be marginally\nslower because of the extra function call.\n\nBut in principle there may be occasions when a \"mixed type\" operator\nwill give a better/more accurate result than doing a type conversion and\nthen a \"uniform type\" operation. So we should probably plan on leaving\nmost of the mixed-type operators and functions in the system.\n\n - Tom\n",
"msg_date": "Thu, 22 Oct 1998 06:10:21 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "\nOn Thu, 22 Oct 1998, Thomas G. Lockhart wrote:\n\n> > While it will be possible, I would like people to concentrate on the\n> > existing 6.4 issues we have, and discourage parallel development on\n> > separate final and features source trees.\n> \n> As Bruce describes, a \"quiet time\" just before and up to a month after a\n> major release is a _really_ good idea. For a couple of reasons:\n> \n> 1) it lets us put out bugfix releases with a minimum of trouble\n> \n> 2) it lets people planning larger changes for the next release to work\n> on a stable, quiet tree, with some assurance that they will likely be\n> able to remerge their work when things unfreeze.\n\n\tI don't agree...the problem is that our times between releases\ntends to be...erratic. It was supposed to be 3+1mos, and is turning into,\nwhat, 5+1? \n\n\tNot all changes to the source tree require a dump/reload to take\neffect...we have 4 patches in the ftp site right now, to v6.3.2, with the\nfirst dated Apr21st, and the last dated Jul30th. With a STABLE vs CURRENT\nbranch, those patches could have been applied to the STABLE branch, and a\nquick v6.3.x release could have been regress tested and released. That in\nitself would have saved some of the problems where ppl had downloaded the\n'newest release' but had problems because they didn't grab the patches to\ngo along with it.\n\n\tThe problem right now is we look at it as being one stream...right\nnow, the only thing that should be left for v6.4 is bug fixes, and there\nshould be no reason to hold up continued development while we wait for\neach possible bug report to flow in.\n\n\tWith two branches, there should be no reason why a patch that\nVadim comes up with to fix a \"rarely hit, but often disastrous\" bug in\nindexes can't be applied and tested in both tree. Then a quick v6.4.1 can\nbe released that those not wishing to run \"latest and greatest\" can run\nwithout having to lose out on that major fix...\n\n\tThe idea is that with very little work on anyone's part, we can\neasily provide a more stable foundation for those starting out and wishing\nto use it in a production environment. Right now, we have a v6.3.2 from\nApril 19th, with four patches that can be applied...but how many ppl would\nactually apply those patches in a production environment? Most ppl would\ndownload and upgrade to v6.3.3 which had those patches applied...\n\n\tI don't know...its something we just did with INN, cause we still\nhad some bugs to work out on 2.2, but some of the developers who don't\nwork on the areas involved are getting restless to get some work\ndone...gives them a chance to move forward without affecting the RELEASE\nscheduale...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 22 Oct 1998 09:05:39 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "> > Thomas, can you check the pg_proc table, and remove any of the \n> > functions that are no longer needed. Seems we have a lot of them in \n> > there that used to do conversions, that are no longer needed.\n> > There are even functions to convert from int2 to int2. We are also\n> > running low on systgem oids.\n> \n> Is this a problem for v6.4? Perhaps we can go with what we have for now,\n> and do a cleanup early in v6.5 development?\n> \n> One thing: I've been thinking that, although we now do \"automatic\n> coersions\" of types which will try to do the right thing by converting\n> expression or function arguments to the appropriate type if there is not\n> direct match in the catalog, a direct match is always/usually better to\n> have. So, for example, there is an operator defined to multiply an int2\n> by an int4. If one actually had an expression that was\n> \n> '2'::int2 * '4'::int4\n> \n> then this operator would be used. If one removed this operator, then the\n> expression would still be evaluated, but done as\n> \n> int4('2'::int2) * '4'::int4\n> \n> which in this case will give the correct result, though be marginally\n> slower because of the extra function call.\n> \n> But in principle there may be occasions when a \"mixed type\" operator\n> will give a better/more accurate result than doing a type conversion and\n> then a \"uniform type\" operation. So we should probably plan on leaving\n> most of the mixed-type operators and functions in the system.\n\nYes. Leave it for 6.4. Just keep it mind next time you are looking in\nthere.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 09:08:08 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "> \tI don't agree...the problem is that our times between releases\n> tends to be...erratic. It was supposed to be 3+1mos, and is turning into,\n> what, 5+1? \n> \n> \tNot all changes to the source tree require a dump/reload to take\n> effect...we have 4 patches in the ftp site right now, to v6.3.2, with the\n> first dated Apr21st, and the last dated Jul30th. With a STABLE vs CURRENT\n> branch, those patches could have been applied to the STABLE branch, and a\n> quick v6.3.x release could have been regress tested and released. That in\n> itself would have saved some of the problems where ppl had downloaded the\n> 'newest release' but had problems because they didn't grab the patches to\n> go along with it.\n> \n> \tThe problem right now is we look at it as being one stream...right\n> now, the only thing that should be left for v6.4 is bug fixes, and there\n> should be no reason to hold up continued development while we wait for\n> each possible bug report to flow in.\n> \n> \tWith two branches, there should be no reason why a patch that\n> Vadim comes up with to fix a \"rarely hit, but often disastrous\" bug in\n> indexes can't be applied and tested in both tree. Then a quick v6.4.1 can\n> be released that those not wishing to run \"latest and greatest\" can run\n> without having to lose out on that major fix...\n> \n> \tThe idea is that with very little work on anyone's part, we can\n> easily provide a more stable foundation for those starting out and wishing\n> to use it in a production environment. Right now, we have a v6.3.2 from\n> April 19th, with four patches that can be applied...but how many ppl would\n> actually apply those patches in a production environment? Most ppl would\n> download and upgrade to v6.3.3 which had those patches applied...\n> \n> \tI don't know...its something we just did with INN, cause we still\n> had some bugs to work out on 2.2, but some of the developers who don't\n> work on the areas involved are getting restless to get some work\n> done...gives them a chance to move forward without affecting the RELEASE\n> scheduale...\n\nWe have this discussion just before every release. The farther away the\nmemory of the 1.* release mess, the less likely Marc is to agree.\n\nIt will not work. It will not provide the stable released we need. \nBut, if you want to Linux it, go ahead.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 09:45:56 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "Thus spake Bruce Momjian\n> It will not work. It will not provide the stable released we need. \n> But, if you want to Linux it, go ahead.\n\nWay to hit Marc's hot buttons, Bruce. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Thu, 22 Oct 1998 10:17:19 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "On Thu, 22 Oct 1998, Bruce Momjian wrote:\n\n> \n> We have this discussion just before every release. The farther away the\n> memory of the 1.* release mess, the less likely Marc is to agree.\n> \n> It will not work. It will not provide the stable released we need. \n> But, if you want to Linux it, go ahead.\n\n\tActually, its a BSDism, not a Linuxism...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 22 Oct 1998 11:26:23 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "On Thu, 22 Oct 1998, D'Arcy J.M. Cain wrote:\n\n> Thus spake Bruce Momjian\n> > It will not work. It will not provide the stable released we need. \n> > But, if you want to Linux it, go ahead.\n> \n> Way to hit Marc's hot buttons, Bruce. :-)\n\n\tNaw...Bruce and I have known each other for much too long for me\nto let that happen :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 22 Oct 1998 11:27:16 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "Thus spake The Hermit Hacker\n> > Way to hit Marc's hot buttons, Bruce. :-)\n> \n> \tNaw...Bruce and I have known each other for much too long for me\n> to let that happen :)\n\nI know. I assumed that that's how he knew your hot button so well. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Thu, 22 Oct 1998 13:06:35 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "On Thu, 22 Oct 1998, D'Arcy J.M. Cain wrote:\n\n> Thus spake The Hermit Hacker\n> > > Way to hit Marc's hot buttons, Bruce. :-)\n> > \n> > \tNaw...Bruce and I have known each other for much too long for me\n> > to let that happen :)\n> \n> I know. I assumed that that's how he knew your hot button so well. :-)\n\n\tGood point :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 22 Oct 1998 14:26:30 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
},
{
"msg_contents": "> I know. I assumed that that's how he knew your hot button so well. :-)\n\nHeeeyyyyy. Waaait a minute. I thought Bruce was giving Mark a\ncompliment...\n\n - Tom\n",
"msg_date": "Thu, 22 Oct 1998 17:34:28 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] CVS Branch Tagging..."
}
] |
[
{
"msg_contents": "I see there are at least other two options for SET/SHOW/RESET commands:\n- KSQO\n- QUERY_LIMIT\n\nI would to underline the help is not updated about KSQO:\n\nhygea=> \\h show\nCommand: show\nDescription: show current run-time environment\nSyntax:\n SHOW DateStyle|GEQO|R_PLANS|QUERY_LIMIT\n\nhygea=> \\h reset\nCommand: reset\nDescription: set run-time environment back to default\nSyntax:\n RESET DateStyle|GEQO|R_PLANS|QUERY_LIMIT\n\nhygea=> \\h set\nCommand: set\nDescription: set run-time environment\nSyntax:\n SET DateStyle TO\n'ISO'|'SQL'|'Postgres'|'European'|'US'|'NonEuropean'\nset GEQO TO 'ON[=#]'|'OFF'\nset R_PLANS TO 'ON'| 'OFF'\nset QUERY_LIMIT TO #\n\nJose'\n",
"msg_date": "Thu, 22 Oct 1998 10:42:55 +0200",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "psql's help"
},
{
"msg_contents": "> I see there are at least other two options for SET/SHOW/RESET commands:\n> - KSQO\n> - QUERY_LIMIT\n> \n> I would to underline the help is not updated about KSQO:\n> \n> hygea=> \\h show\n> Command: show\n> Description: show current run-time environment\n> Syntax:\n> SHOW DateStyle|GEQO|R_PLANS|QUERY_LIMIT\n> \n> hygea=> \\h reset\n> Command: reset\n> Description: set run-time environment back to default\n> Syntax:\n> RESET DateStyle|GEQO|R_PLANS|QUERY_LIMIT\n> \n> hygea=> \\h set\n> Command: set\n> Description: set run-time environment\n> Syntax:\n> SET DateStyle TO\n> 'ISO'|'SQL'|'Postgres'|'European'|'US'|'NonEuropean'\n> set GEQO TO 'ON[=#]'|'OFF'\n> set R_PLANS TO 'ON'| 'OFF'\n> set QUERY_LIMIT TO #\n\nThis is because KQSO is a workaround, that may even be gone by 6.4\nfinal.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 09:10:49 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help"
},
{
"msg_contents": "> > SET DateStyle TO\n> > 'ISO'|'SQL'|'Postgres'|'European'|'US'|'NonEuropean'\n> > set GEQO TO 'ON[=#]'|'OFF'\n> > set R_PLANS TO 'ON'| 'OFF'\n> > set QUERY_LIMIT TO #\n> \n> This is because KQSO is a workaround, that may even be gone by 6.4\n> final.\n\n I hope the QUERY_LIMIT too. \n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n",
"msg_date": "Thu, 22 Oct 1998 15:34:21 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help"
},
{
"msg_contents": "> > > SET DateStyle TO\n> > > 'ISO'|'SQL'|'Postgres'|'European'|'US'|'NonEuropean'\n> > > set GEQO TO 'ON[=#]'|'OFF'\n> > > set R_PLANS TO 'ON'| 'OFF'\n> > > set QUERY_LIMIT TO #\n> > \n> > This is because KQSO is a workaround, that may even be gone by 6.4\n> > final.\n> \n> I hope the QUERY_LIMIT too. \n\nI still have that cnfify() possible fix to review for KQSO. Are you\nstill thinking limit for 6.4 final, or a minor release after that?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 09:51:19 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help"
},
{
"msg_contents": "> > >\n> > > I hope the QUERY_LIMIT too.\n> >\n> > I still have that cnfify() possible fix to review for KQSO. Are you\n> > still thinking limit for 6.4 final, or a minor release after that?\n> \n> I posted the part that is the minimum applied to 6.4 to make\n> adding LIMIT later non-initdb earlier. Anyway, here it's\n> again.\n\nAlready applied. I assume it is the same as the one I applied.\n\n> \n> My LIMIT implementation that does it like the SET in the\n> toplevel executor (but via parsetree values) is ready for\n> production. I only held it back because it's feature, not\n> bugfix.\n> \n> Do you want it in 6.4 final?\n\nWe are close to final, and can easily put it in 6.4.1, which I am sure\nwe will need, and if we split CVS trees, you'll have lots of minor\nversions to pick from. :-)\n\nSeems like it would be a nice minor release item, but the problem is\nthat minor releases aren't tested as much as major ones. How confident\nare you in the code? What do others thing?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 10:24:08 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help (the LIMIT stuff)"
},
{
"msg_contents": "> >\n> > I hope the QUERY_LIMIT too.\n>\n> I still have that cnfify() possible fix to review for KQSO. Are you\n> still thinking limit for 6.4 final, or a minor release after that?\n\n I posted the part that is the minimum applied to 6.4 to make\n adding LIMIT later non-initdb earlier. Anyway, here it's\n again.\n\n My LIMIT implementation that does it like the SET in the\n toplevel executor (but via parsetree values) is ready for\n production. I only held it back because it's feature, not\n bugfix.\n\n Do you want it in 6.4 final?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\ndiff -cr src.orig/backend/nodes/copyfuncs.c src/backend/nodes/copyfuncs.c\n*** src.orig/backend/nodes/copyfuncs.c\tFri Oct 16 11:53:40 1998\n--- src/backend/nodes/copyfuncs.c\tFri Oct 16 13:32:35 1998\n***************\n*** 1578,1583 ****\n--- 1578,1586 ----\n \t\tnewnode->unionClause = temp_list;\n \t}\n \n+ \tNode_Copy(from, newnode, limitOffset);\n+ \tNode_Copy(from, newnode, limitCount);\n+ \n \treturn newnode;\n }\n \ndiff -cr src.orig/backend/nodes/outfuncs.c src/backend/nodes/outfuncs.c\n*** src.orig/backend/nodes/outfuncs.c\tFri Oct 16 11:53:40 1998\n--- src/backend/nodes/outfuncs.c\tFri Oct 16 13:30:50 1998\n***************\n*** 259,264 ****\n--- 259,268 ----\n \tappendStringInfo(str, (node->hasSubLinks ? \"true\" : \"false\"));\n \tappendStringInfo(str, \" :unionClause \");\n \t_outNode(str, node->unionClause);\n+ \tappendStringInfo(str, \" :limitOffset \");\n+ \t_outNode(str, node->limitOffset);\n+ \tappendStringInfo(str, \" :limitCount \");\n+ \t_outNode(str, node->limitCount);\n }\n \n static void\ndiff -cr src.orig/backend/nodes/readfuncs.c src/backend/nodes/readfuncs.c\n*** src.orig/backend/nodes/readfuncs.c\tFri Oct 16 11:53:40 1998\n--- src/backend/nodes/readfuncs.c\tFri Oct 16 13:31:43 1998\n***************\n*** 163,168 ****\n--- 163,174 ----\n \ttoken = lsptok(NULL, &length);\t\t/* skip :unionClause */\n \tlocal_node->unionClause = nodeRead(true);\n \n+ \ttoken = lsptok(NULL, &length);\t\t/* skip :limitOffset */\n+ \tlocal_node->limitOffset = nodeRead(true);\n+ \n+ \ttoken = lsptok(NULL, &length);\t\t/* skip :limitCount */\n+ \tlocal_node->limitCount = nodeRead(true);\n+ \n \treturn local_node;\n }\n \ndiff -cr src.orig/include/nodes/parsenodes.h src/include/nodes/parsenodes.h\n*** src.orig/include/nodes/parsenodes.h\tFri Oct 16 11:53:58 1998\n--- src/include/nodes/parsenodes.h\tFri Oct 16 13:35:32 1998\n***************\n*** 60,65 ****\n--- 60,67 ----\n \n \tList\t *unionClause;\t/* unions are linked under the previous\n \t\t\t\t\t\t\t\t * query */\n+ \tNode\t *limitOffset;\t/* # of result tuples to skip */\n+ \tNode\t *limitCount;\t\t/* # of result tuples to return */\n \n \t/* internal to planner */\n \tList\t *base_rel_list;\t/* base relation list */\n***************\n*** 639,644 ****\n--- 641,648 ----\n \tchar\t *portalname;\t\t/* the portal (cursor) to create */\n \tbool\t\tbinary;\t\t\t/* a binary (internal) portal? */\n \tbool\t\tunionall;\t\t/* union without unique sort */\n+ \tNode\t *limitOffset;\t/* # of result tuples to skip */\n+ \tNode\t *limitCount;\t\t/* # of result tuples to return */\n } SelectStmt;\n \n \n",
"msg_date": "Thu, 22 Oct 1998 16:33:50 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help (the LIMIT stuff)"
},
{
"msg_contents": ">\n> > > >\n> > > > I hope the QUERY_LIMIT too.\n> > >\n> > > I still have that cnfify() possible fix to review for KQSO. Are you\n> > > still thinking limit for 6.4 final, or a minor release after that?\n> >\n> > I posted the part that is the minimum applied to 6.4 to make\n> > adding LIMIT later non-initdb earlier. Anyway, here it's\n> > again.\n>\n> Already applied. I assume it is the same as the one I applied.\n\n Seen, thanks. Your 'Applied' just arrived after I packed it\n again. It's the same.\n\n> We are close to final, and can easily put it in 6.4.1, which I am sure\n> we will need, and if we split CVS trees, you'll have lots of minor\n> versions to pick from. :-)\n>\n> Seems like it would be a nice minor release item, but the problem is\n> that minor releases aren't tested as much as major ones. How confident\n> are you in the code? What do others thing?\n\n I regression tested it, and did additional tests in the\n SPI/PL area. It works. It only touches the parser and the\n executor. Rules, planner/optimizer just bypass the values in\n the parsetree. The parser and the executor are parts of\n Postgres I feel very familiar with (not so in the optimizer).\n I trust in the code and would use it in a production\n environment.\n\n It's below.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\ndiff -cr src.orig/backend/commands/command.c src/backend/commands/command.c\n*** src.orig/backend/commands/command.c\tFri Oct 16 11:53:38 1998\n--- src/backend/commands/command.c\tFri Oct 16 12:56:44 1998\n***************\n*** 39,44 ****\n--- 39,45 ----\n #include \"utils/mcxt.h\"\n #include \"utils/portal.h\"\n #include \"utils/syscache.h\"\n+ #include \"string.h\"\n \n /* ----------------\n *\t\tPortalExecutorHeapMemory stuff\n***************\n*** 101,106 ****\n--- 102,108 ----\n \tint\t\t\tfeature;\n \tQueryDesc *queryDesc;\n \tMemoryContext context;\n+ \tConst\t\tlimcount;\n \n \t/* ----------------\n \t *\tsanity checks\n***************\n*** 113,118 ****\n--- 115,134 ----\n \t}\n \n \t/* ----------------\n+ \t *\tCreate a const node from the given count value\n+ \t * ----------------\n+ \t */\n+ \tmemset(&limcount, 0, sizeof(limcount));\n+ \tlimcount.type\t\t= T_Const;\n+ \tlimcount.consttype\t= INT4OID;\n+ \tlimcount.constlen\t= sizeof(int4);\n+ \tlimcount.constvalue\t= (Datum)count;\n+ \tlimcount.constisnull\t= FALSE;\n+ \tlimcount.constbyval\t= TRUE;\n+ \tlimcount.constisset\t= FALSE;\n+ \tlimcount.constiscast\t= FALSE;\n+ \n+ \t/* ----------------\n \t *\tget the portal from the portal name\n \t * ----------------\n \t */\n***************\n*** 176,182 ****\n \tPortalExecutorHeapMemory = (MemoryContext)\n \t\tPortalGetHeapMemory(portal);\n \n! \tExecutorRun(queryDesc, PortalGetState(portal), feature, count);\n \n \tif (dest == None)\t\t\t/* MOVE */\n \t\tpfree(queryDesc);\n--- 192,198 ----\n \tPortalExecutorHeapMemory = (MemoryContext)\n \t\tPortalGetHeapMemory(portal);\n \n! \tExecutorRun(queryDesc, PortalGetState(portal), feature, (Node *)NULL, (Node *)&limcount);\n \n \tif (dest == None)\t\t\t/* MOVE */\n \t\tpfree(queryDesc);\ndiff -cr src.orig/backend/executor/execMain.c src/backend/executor/execMain.c\n*** src.orig/backend/executor/execMain.c\tFri Oct 16 11:53:38 1998\n--- src/backend/executor/execMain.c\tFri Oct 16 20:05:19 1998\n***************\n*** 64,69 ****\n--- 64,70 ----\n static void EndPlan(Plan *plan, EState *estate);\n static TupleTableSlot *ExecutePlan(EState *estate, Plan *plan,\n \t\t\tQuery *parseTree, CmdType operation,\n+ \t\t\tint offsetTuples,\n \t\t\tint numberTuples, ScanDirection direction,\n \t\t\tvoid (*printfunc) ());\n static void ExecRetrieve(TupleTableSlot *slot, void (*printfunc) (),\n***************\n*** 163,169 ****\n * ----------------------------------------------------------------\n */\n TupleTableSlot *\n! ExecutorRun(QueryDesc *queryDesc, EState *estate, int feature, int count)\n {\n \tCmdType\t\toperation;\n \tQuery\t *parseTree;\n--- 164,170 ----\n * ----------------------------------------------------------------\n */\n TupleTableSlot *\n! ExecutorRun(QueryDesc *queryDesc, EState *estate, int feature, Node *limoffset, Node *limcount)\n {\n \tCmdType\t\toperation;\n \tQuery\t *parseTree;\n***************\n*** 171,176 ****\n--- 172,179 ----\n \tTupleTableSlot *result;\n \tCommandDest dest;\n \tvoid\t\t(*destination) ();\n+ \tint\t\toffset = 0;\n+ \tint\t\tcount = 0;\n \n \t/******************\n \t *\tsanity checks\n***************\n*** 191,196 ****\n--- 194,289 ----\n \testate->es_processed = 0;\n \testate->es_lastoid = InvalidOid;\n \n+ \t/******************\n+ \t *\tif given get the offset of the LIMIT clause\n+ \t ******************\n+ \t */\n+ \tif (limoffset != NULL)\n+ \t{\n+ \t\tConst\t\t*coffset;\n+ \t\tParam\t\t*poffset;\n+ \t\tParamListInfo\tparamLI;\n+ \t\tint\t\ti;\n+ \n+ \t\tswitch (nodeTag(limoffset))\n+ \t\t{\n+ \t\t\tcase T_Const:\n+ \t\t\t\tcoffset = (Const *)limoffset;\n+ \t\t\t\toffset = (int)(coffset->constvalue);\n+ \t\t\t\tbreak;\n+ \n+ \t\t\tcase T_Param:\n+ \t\t\t\tpoffset = (Param *)limoffset;\n+ \t\t\t\tparamLI = estate->es_param_list_info;\n+ \n+ \t\t\t\tif (paramLI == NULL)\n+ \t\t\t\t\telog(ERROR, \"parameter for limit offset not in executor state\");\n+ \t\t\t\tfor (i = 0; paramLI[i].kind != PARAM_INVALID; i++)\n+ \t\t\t\t{\n+ \t\t\t\t\tif (paramLI[i].kind == PARAM_NUM && paramLI[i].id == poffset->paramid)\n+ \t\t\t\t\t\tbreak;\n+ \t\t\t\t}\n+ \t\t\t\tif (paramLI[i].kind == PARAM_INVALID)\n+ \t\t\t\t\telog(ERROR, \"parameter for limit offset not in executor state\");\n+ \t\t\t\tif (paramLI[i].isnull)\n+ \t\t\t\t\telog(ERROR, \"limit offset cannot be NULL value\");\n+ \t\t\t\toffset = (int)(paramLI[i].value);\n+ \n+ \t\t\t\tbreak;\n+ \n+ \t\t\tdefault:\n+ \t\t\t\telog(ERROR, \"unexpected node type %d as limit offset\", nodeTag(limoffset));\n+ \t\t}\n+ \n+ \t\tif (offset < 0)\n+ \t\t\telog(ERROR, \"limit offset cannot be negative\");\n+ \t}\n+ \n+ \t/******************\n+ \t *\tif given get the count of the LIMIT clause\n+ \t ******************\n+ \t */\n+ \tif (limcount != NULL)\n+ \t{\n+ \t\tConst\t\t*ccount;\n+ \t\tParam\t\t*pcount;\n+ \t\tParamListInfo\tparamLI;\n+ \t\tint\t\ti;\n+ \n+ \t\tswitch (nodeTag(limcount))\n+ \t\t{\n+ \t\t\tcase T_Const:\n+ \t\t\t\tccount = (Const *)limcount;\n+ \t\t\t\tcount = (int)(ccount->constvalue);\n+ \t\t\t\tbreak;\n+ \n+ \t\t\tcase T_Param:\n+ \t\t\t\tpcount = (Param *)limcount;\n+ \t\t\t\tparamLI = estate->es_param_list_info;\n+ \n+ \t\t\t\tif (paramLI == NULL)\n+ \t\t\t\t\telog(ERROR, \"parameter for limit count not in executor state\");\n+ \t\t\t\tfor (i = 0; paramLI[i].kind != PARAM_INVALID; i++)\n+ \t\t\t\t{\n+ \t\t\t\t\tif (paramLI[i].kind == PARAM_NUM && paramLI[i].id == pcount->paramid)\n+ \t\t\t\t\t\tbreak;\n+ \t\t\t\t}\n+ \t\t\t\tif (paramLI[i].kind == PARAM_INVALID)\n+ \t\t\t\t\telog(ERROR, \"parameter for limit count not in executor state\");\n+ \t\t\t\tif (paramLI[i].isnull)\n+ \t\t\t\t\telog(ERROR, \"limit count cannot be NULL value\");\n+ \t\t\t\tcount = (int)(paramLI[i].value);\n+ \n+ \t\t\t\tbreak;\n+ \n+ \t\t\tdefault:\n+ \t\t\t\telog(ERROR, \"unexpected node type %d as limit count\", nodeTag(limcount));\n+ \t\t}\n+ \n+ \t\tif (count < 0)\n+ \t\t\telog(ERROR, \"limit count cannot be negative\");\n+ \t}\n+ \n \tswitch (feature)\n \t{\n \n***************\n*** 199,205 ****\n \t\t\t\t\t\t\t\t plan,\n \t\t\t\t\t\t\t\t parseTree,\n \t\t\t\t\t\t\t\t operation,\n! \t\t\t\t\t\t\t\t ALL_TUPLES,\n \t\t\t\t\t\t\t\t ForwardScanDirection,\n \t\t\t\t\t\t\t\t destination);\n \t\t\tbreak;\n--- 292,299 ----\n \t\t\t\t\t\t\t\t plan,\n \t\t\t\t\t\t\t\t parseTree,\n \t\t\t\t\t\t\t\t operation,\n! \t\t\t\t\t\t\t\t offset,\n! \t\t\t\t\t\t\t\t count,\n \t\t\t\t\t\t\t\t ForwardScanDirection,\n \t\t\t\t\t\t\t\t destination);\n \t\t\tbreak;\n***************\n*** 208,213 ****\n--- 302,308 ----\n \t\t\t\t\t\t\t\t plan,\n \t\t\t\t\t\t\t\t parseTree,\n \t\t\t\t\t\t\t\t operation,\n+ \t\t\t\t\t\t\t\t offset,\n \t\t\t\t\t\t\t\t count,\n \t\t\t\t\t\t\t\t ForwardScanDirection,\n \t\t\t\t\t\t\t\t destination);\n***************\n*** 222,227 ****\n--- 317,323 ----\n \t\t\t\t\t\t\t\t plan,\n \t\t\t\t\t\t\t\t parseTree,\n \t\t\t\t\t\t\t\t operation,\n+ \t\t\t\t\t\t\t\t offset,\n \t\t\t\t\t\t\t\t count,\n \t\t\t\t\t\t\t\t BackwardScanDirection,\n \t\t\t\t\t\t\t\t destination);\n***************\n*** 237,242 ****\n--- 333,339 ----\n \t\t\t\t\t\t\t\t plan,\n \t\t\t\t\t\t\t\t parseTree,\n \t\t\t\t\t\t\t\t operation,\n+ \t\t\t\t\t\t\t\t 0,\n \t\t\t\t\t\t\t\t ONE_TUPLE,\n \t\t\t\t\t\t\t\t ForwardScanDirection,\n \t\t\t\t\t\t\t\t destination);\n***************\n*** 691,696 ****\n--- 788,794 ----\n \t\t\tPlan *plan,\n \t\t\tQuery *parseTree,\n \t\t\tCmdType operation,\n+ \t\t\tint offsetTuples,\n \t\t\tint numberTuples,\n \t\t\tScanDirection direction,\n \t\t\tvoid (*printfunc) ())\n***************\n*** 742,747 ****\n--- 840,859 ----\n \t\t{\n \t\t\tresult = NULL;\n \t\t\tbreak;\n+ \t\t}\n+ \n+ \t\t/******************\n+ \t\t *\tFor now we completely execute the plan and skip\n+ \t\t *\tresult tuples if requested by LIMIT offset.\n+ \t\t *\tFinally we should try to do it in deeper levels\n+ \t\t *\tif possible (during index scan)\n+ \t\t *\t- Jan\n+ \t\t ******************\n+ \t\t */\n+ \t\tif (offsetTuples > 0)\n+ \t\t{\n+ \t\t\t--offsetTuples;\n+ \t\t\tcontinue;\n \t\t}\n \n \t\t/******************\ndiff -cr src.orig/backend/executor/functions.c src/backend/executor/functions.c\n*** src.orig/backend/executor/functions.c\tFri Oct 16 11:53:38 1998\n--- src/backend/executor/functions.c\tFri Oct 16 19:01:02 1998\n***************\n*** 130,135 ****\n--- 130,138 ----\n \t\t\t\t\t\t\t\t\t None);\n \t\testate = CreateExecutorState();\n \n+ \t\tif (queryTree->limitOffset != NULL || queryTree->limitCount != NULL)\n+ \t\t\telog(ERROR, \"LIMIT clause from SQL functions not yet implemented\");\n+ \n \t\tif (nargs > 0)\n \t\t{\n \t\t\tint\t\t\ti;\n***************\n*** 200,206 ****\n \n \tfeature = (LAST_POSTQUEL_COMMAND(es)) ? EXEC_RETONE : EXEC_RUN;\n \n! \treturn ExecutorRun(es->qd, es->estate, feature, 0);\n }\n \n static void\n--- 203,209 ----\n \n \tfeature = (LAST_POSTQUEL_COMMAND(es)) ? EXEC_RETONE : EXEC_RUN;\n \n! \treturn ExecutorRun(es->qd, es->estate, feature, (Node *)NULL, (Node *)NULL);\n }\n \n static void\ndiff -cr src.orig/backend/executor/spi.c src/backend/executor/spi.c\n*** src.orig/backend/executor/spi.c\tFri Oct 16 11:53:39 1998\n--- src/backend/executor/spi.c\tFri Oct 16 19:25:33 1998\n***************\n*** 791,796 ****\n--- 791,798 ----\n \tbool\t\tisRetrieveIntoRelation = false;\n \tchar\t *intoName = NULL;\n \tint\t\t\tres;\n+ \tConst\t\ttcount_const;\n+ \tNode\t\t*count = NULL;\n \n \tswitch (operation)\n \t{\n***************\n*** 825,830 ****\n--- 827,865 ----\n \t\t\treturn SPI_ERROR_OPUNKNOWN;\n \t}\n \n+ \t/* ----------------\n+ \t *\tGet the query LIMIT tuple count\n+ \t * ----------------\n+ \t */\n+ \tif (parseTree->limitCount != NULL)\n+ \t{\n+ \t\t/* ----------------\n+ \t\t * A limit clause in the parsetree overrides the\n+ \t\t *\ttcount parameter\n+ \t\t * ----------------\n+ \t\t */\n+ \t\tcount = parseTree->limitCount;\n+ \t}\n+ \telse\n+ \t{\n+ \t\t/* ----------------\n+ \t\t * No LIMIT clause in parsetree. Use a local Const node\n+ \t\t *\tto put tcount into it\n+ \t\t * ----------------\n+ \t\t */\n+ \t\tmemset(&tcount_const, 0, sizeof(tcount_const));\n+ \t\ttcount_const.type = T_Const;\n+ \t\ttcount_const.consttype = INT4OID;\n+ \t\ttcount_const.constlen = sizeof(int4);\n+ \t\ttcount_const.constvalue = (Datum)tcount;\n+ \t\ttcount_const.constisnull = FALSE;\n+ \t\ttcount_const.constbyval = TRUE;\n+ \t\ttcount_const.constisset = FALSE;\n+ \t\ttcount_const.constiscast = FALSE;\n+ \n+ \t\tcount = (Node *)&tcount_const;\n+ \t}\n+ \n \tif (state == NULL)\t\t\t/* plan preparation */\n \t\treturn res;\n #ifdef SPI_EXECUTOR_STATS\n***************\n*** 845,851 ****\n \t\treturn SPI_OK_CURSOR;\n \t}\n \n! \tExecutorRun(queryDesc, state, EXEC_FOR, tcount);\n \n \t_SPI_current->processed = state->es_processed;\n \tif (operation == CMD_SELECT && queryDesc->dest == SPI)\n--- 880,886 ----\n \t\treturn SPI_OK_CURSOR;\n \t}\n \n! \tExecutorRun(queryDesc, state, EXEC_FOR, parseTree->limitOffset, count);\n \n \t_SPI_current->processed = state->es_processed;\n \tif (operation == CMD_SELECT && queryDesc->dest == SPI)\ndiff -cr src.orig/backend/parser/analyze.c src/backend/parser/analyze.c\n*** src.orig/backend/parser/analyze.c\tFri Oct 16 11:53:41 1998\n--- src/backend/parser/analyze.c\tFri Oct 16 13:29:27 1998\n***************\n*** 180,186 ****\n--- 180,190 ----\n \n \t\tcase T_SelectStmt:\n \t\t\tif (!((SelectStmt *) parseTree)->portalname)\n+ \t\t\t{\n \t\t\t\tresult = transformSelectStmt(pstate, (SelectStmt *) parseTree);\n+ \t\t\t\tresult->limitOffset = ((SelectStmt *)parseTree)->limitOffset;\n+ \t\t\t\tresult->limitCount = ((SelectStmt *)parseTree)->limitCount;\n+ \t\t\t}\n \t\t\telse\n \t\t\t\tresult = transformCursorStmt(pstate, (SelectStmt *) parseTree);\n \t\t\tbreak;\ndiff -cr src.orig/backend/parser/gram.y src/backend/parser/gram.y\n*** src.orig/backend/parser/gram.y\tFri Oct 16 11:53:42 1998\n--- src/backend/parser/gram.y\tSun Oct 18 22:20:36 1998\n***************\n*** 45,50 ****\n--- 45,51 ----\n #include \"catalog/catname.h\"\n #include \"utils/elog.h\"\n #include \"access/xact.h\"\n+ #include \"catalog/pg_type.h\"\n \n #ifdef MULTIBYTE\n #include \"mb/pg_wchar.h\"\n***************\n*** 163,169 ****\n \t\tsort_clause, sortby_list, index_params, index_list, name_list,\n \t\tfrom_clause, from_list, opt_array_bounds, nest_array_bounds,\n \t\texpr_list, attrs, res_target_list, res_target_list2,\n! \t\tdef_list, opt_indirection, group_clause, TriggerFuncArgs\n \n %type <node>\tfunc_return\n %type <boolean>\tset_opt\n--- 164,171 ----\n \t\tsort_clause, sortby_list, index_params, index_list, name_list,\n \t\tfrom_clause, from_list, opt_array_bounds, nest_array_bounds,\n \t\texpr_list, attrs, res_target_list, res_target_list2,\n! \t\tdef_list, opt_indirection, group_clause, TriggerFuncArgs,\n! \t\topt_select_limit\n \n %type <node>\tfunc_return\n %type <boolean>\tset_opt\n***************\n*** 192,197 ****\n--- 194,201 ----\n \n %type <ival>\tfetch_how_many\n \n+ %type <node>\tselect_limit_value select_offset_value\n+ \n %type <list>\tOptSeqList\n %type <defelt>\tOptSeqElem\n \n***************\n*** 267,273 ****\n \t\tFALSE_P, FETCH, FLOAT, FOR, FOREIGN, FROM, FULL,\n \t\tGRANT, GROUP, HAVING, HOUR_P,\n \t\tIN, INNER_P, INSENSITIVE, INSERT, INTERVAL, INTO, IS,\n! \t\tJOIN, KEY, LANGUAGE, LEADING, LEFT, LIKE, LOCAL,\n \t\tMATCH, MINUTE_P, MONTH_P, NAMES,\n \t\tNATIONAL, NATURAL, NCHAR, NEXT, NO, NOT, NULL_P, NUMERIC,\n \t\tOF, ON, ONLY, OPTION, OR, ORDER, OUTER_P,\n--- 271,277 ----\n \t\tFALSE_P, FETCH, FLOAT, FOR, FOREIGN, FROM, FULL,\n \t\tGRANT, GROUP, HAVING, HOUR_P,\n \t\tIN, INNER_P, INSENSITIVE, INSERT, INTERVAL, INTO, IS,\n! \t\tJOIN, KEY, LANGUAGE, LEADING, LEFT, LIKE, LIMIT, LOCAL,\n \t\tMATCH, MINUTE_P, MONTH_P, NAMES,\n \t\tNATIONAL, NATURAL, NCHAR, NEXT, NO, NOT, NULL_P, NUMERIC,\n \t\tOF, ON, ONLY, OPTION, OR, ORDER, OUTER_P,\n***************\n*** 299,305 ****\n \t\tINCREMENT, INDEX, INHERITS, INSTEAD, ISNULL,\n \t\tLANCOMPILER, LISTEN, LOAD, LOCATION, LOCK_P, MAXVALUE, MINVALUE, MOVE,\n \t\tNEW, NOCREATEDB, NOCREATEUSER, NONE, NOTHING, NOTIFY, NOTNULL,\n! \t\tOIDS, OPERATOR, PASSWORD, PROCEDURAL,\n \t\tRECIPE, RENAME, RESET, RETURNS, ROW, RULE,\n \t\tSEQUENCE, SERIAL, SETOF, SHOW, START, STATEMENT, STDIN, STDOUT, TRUSTED, \n \t\tUNLISTEN, UNTIL, VACUUM, VALID, VERBOSE, VERSION\n--- 303,309 ----\n \t\tINCREMENT, INDEX, INHERITS, INSTEAD, ISNULL,\n \t\tLANCOMPILER, LISTEN, LOAD, LOCATION, LOCK_P, MAXVALUE, MINVALUE, MOVE,\n \t\tNEW, NOCREATEDB, NOCREATEUSER, NONE, NOTHING, NOTIFY, NOTNULL,\n! \t\tOFFSET, OIDS, OPERATOR, PASSWORD, PROCEDURAL,\n \t\tRECIPE, RENAME, RESET, RETURNS, ROW, RULE,\n \t\tSEQUENCE, SERIAL, SETOF, SHOW, START, STATEMENT, STDIN, STDOUT, TRUSTED, \n \t\tUNLISTEN, UNTIL, VACUUM, VALID, VERBOSE, VERSION\n***************\n*** 2591,2596 ****\n--- 2595,2601 ----\n \t\t\t result from_clause where_clause\n \t\t\t group_clause having_clause\n \t\t\t union_clause sort_clause\n+ \t\t\t opt_select_limit\n \t\t\t\t{\n \t\t\t\t\tSelectStmt *n = makeNode(SelectStmt);\n \t\t\t\t\tn->unique = $2;\n***************\n*** 2602,2607 ****\n--- 2607,2622 ----\n \t\t\t\t\tn->havingClause = $8;\n \t\t\t\t\tn->unionClause = $9;\n \t\t\t\t\tn->sortClause = $10;\n+ \t\t\t\t\tif ($11 != NIL)\n+ \t\t\t\t\t{\n+ \t\t\t\t\t\tn->limitOffset = nth(0, $11);\n+ \t\t\t\t\t\tn->limitCount = nth(1, $11);\n+ \t\t\t\t\t}\n+ \t\t\t\t\telse\n+ \t\t\t\t\t{\n+ \t\t\t\t\t\tn->limitOffset = NULL;\n+ \t\t\t\t\t\tn->limitCount = NULL;\n+ \t\t\t\t\t}\n \t\t\t\t\t$$ = (Node *)n;\n \t\t\t\t}\n \t\t;\n***************\n*** 2699,2704 ****\n--- 2714,2794 ----\n \t\t| ASC\t\t\t\t\t\t\t\t\t{ $$ = \"<\"; }\n \t\t| DESC\t\t\t\t\t\t\t\t\t{ $$ = \">\"; }\n \t\t| /*EMPTY*/\t\t\t\t\t\t\t\t{ $$ = \"<\"; /*default*/ }\n+ \t\t;\n+ \n+ opt_select_limit:\tLIMIT select_offset_value ',' select_limit_value\n+ \t{ $$ = lappend(lappend(NIL, $2), $4); }\n+ \t\t| LIMIT select_limit_value OFFSET select_offset_value\n+ \t{ $$ = lappend(lappend(NIL, $4), $2); }\n+ \t\t| LIMIT select_limit_value\n+ \t{ $$ = lappend(lappend(NIL, NULL), $2); }\n+ \t\t| OFFSET select_offset_value LIMIT select_limit_value\n+ \t{ $$ = lappend(lappend(NIL, $2), $4); }\n+ \t\t| OFFSET select_offset_value\n+ \t{ $$ = lappend(lappend(NIL, $2), NULL); }\n+ \t\t| /* EMPTY */\n+ \t{ $$ = NIL; }\n+ \t\t;\n+ \n+ select_limit_value:\tIconst\n+ \t\t\t{\n+ \t\t\t\tConst\t*n = makeNode(Const);\n+ \n+ \t\t\t\tif ($1 < 1)\n+ \t\t\t\t\telog(ERROR, \"selection limit must be ALL or a positive integer value > 0\");\n+ \n+ \t\t\t\tn->consttype\t= INT4OID;\n+ \t\t\t\tn->constlen\t= sizeof(int4);\n+ \t\t\t\tn->constvalue\t= (Datum)$1;\n+ \t\t\t\tn->constisnull\t= FALSE;\n+ \t\t\t\tn->constbyval\t= TRUE;\n+ \t\t\t\tn->constisset\t= FALSE;\n+ \t\t\t\tn->constiscast\t= FALSE;\n+ \t\t\t\t$$ = (Node *)n;\n+ \t\t\t}\n+ \t\t| ALL\n+ \t\t\t{\n+ \t\t\t\tConst\t*n = makeNode(Const);\n+ \t\t\t\tn->consttype\t= INT4OID;\n+ \t\t\t\tn->constlen\t= sizeof(int4);\n+ \t\t\t\tn->constvalue\t= (Datum)0;\n+ \t\t\t\tn->constisnull\t= FALSE;\n+ \t\t\t\tn->constbyval\t= TRUE;\n+ \t\t\t\tn->constisset\t= FALSE;\n+ \t\t\t\tn->constiscast\t= FALSE;\n+ \t\t\t\t$$ = (Node *)n;\n+ \t\t\t}\n+ \t\t| PARAM\n+ \t\t\t{\n+ \t\t\t\tParam\t*n = makeNode(Param);\n+ \t\t\t\tn->paramkind = PARAM_NUM;\n+ \t\t\t\tn->paramid = $1;\n+ \t\t\t\tn->paramtype = INT4OID;\n+ \t\t\t\t$$ = (Node *)n;\n+ \t\t\t}\n+ \t\t;\n+ \n+ select_offset_value:\tIconst\n+ \t\t\t{\n+ \t\t\t\tConst\t*n = makeNode(Const);\n+ \n+ \t\t\t\tn->consttype\t= INT4OID;\n+ \t\t\t\tn->constlen\t= sizeof(int4);\n+ \t\t\t\tn->constvalue\t= (Datum)$1;\n+ \t\t\t\tn->constisnull\t= FALSE;\n+ \t\t\t\tn->constbyval\t= TRUE;\n+ \t\t\t\tn->constisset\t= FALSE;\n+ \t\t\t\tn->constiscast\t= FALSE;\n+ \t\t\t\t$$ = (Node *)n;\n+ \t\t\t}\n+ \t\t| PARAM\n+ \t\t\t{\n+ \t\t\t\tParam\t*n = makeNode(Param);\n+ \t\t\t\tn->paramkind = PARAM_NUM;\n+ \t\t\t\tn->paramid = $1;\n+ \t\t\t\tn->paramtype = INT4OID;\n+ \t\t\t\t$$ = (Node *)n;\n+ \t\t\t}\n \t\t;\n \n /*\ndiff -cr src.orig/backend/parser/keywords.c src/backend/parser/keywords.c\n*** src.orig/backend/parser/keywords.c\tFri Oct 16 11:53:42 1998\n--- src/backend/parser/keywords.c\tSun Oct 18 22:13:29 1998\n***************\n*** 128,133 ****\n--- 128,134 ----\n \t{\"leading\", LEADING},\n \t{\"left\", LEFT},\n \t{\"like\", LIKE},\n+ \t{\"limit\", LIMIT},\n \t{\"listen\", LISTEN},\n \t{\"load\", LOAD},\n \t{\"local\", LOCAL},\n***************\n*** 156,161 ****\n--- 157,163 ----\n \t{\"null\", NULL_P},\n \t{\"numeric\", NUMERIC},\n \t{\"of\", OF},\n+ \t{\"offset\", OFFSET},\n \t{\"oids\", OIDS},\n \t{\"old\", CURRENT},\n \t{\"on\", ON},\ndiff -cr src.orig/backend/rewrite/rewriteDefine.c src/backend/rewrite/rewriteDefine.c\n*** src.orig/backend/rewrite/rewriteDefine.c\tFri Oct 16 11:53:46 1998\n--- src/backend/rewrite/rewriteDefine.c\tFri Oct 16 13:48:55 1998\n***************\n*** 312,317 ****\n--- 312,323 ----\n \t\theap_close(event_relation);\n \n \t\t/*\n+ \t\t * LIMIT in view is not supported\n+ \t\t */\n+ \t\tif (query->limitOffset != NULL || query->limitCount != NULL)\n+ \t\t\telog(ERROR, \"LIMIT clause not supported in views\");\n+ \n+ \t\t/*\n \t\t * ... and finally the rule must be named _RETviewname.\n \t\t */\n \t\tsprintf(expected_name, \"_RET%s\", event_obj->relname);\ndiff -cr src.orig/backend/tcop/pquery.c src/backend/tcop/pquery.c\n*** src.orig/backend/tcop/pquery.c\tFri Oct 16 11:53:47 1998\n--- src/backend/tcop/pquery.c\tFri Oct 16 14:02:36 1998\n***************\n*** 40,46 ****\n #include \"commands/command.h\"\n \n static char *CreateOperationTag(int operationType);\n! static void ProcessQueryDesc(QueryDesc *queryDesc);\n \n \n /* ----------------------------------------------------------------\n--- 40,46 ----\n #include \"commands/command.h\"\n \n static char *CreateOperationTag(int operationType);\n! static void ProcessQueryDesc(QueryDesc *queryDesc, Node *limoffset, Node *limcount);\n \n \n /* ----------------------------------------------------------------\n***************\n*** 205,211 ****\n * ----------------------------------------------------------------\n */\n static void\n! ProcessQueryDesc(QueryDesc *queryDesc)\n {\n \tQuery\t *parseTree;\n \tPlan\t *plan;\n--- 205,211 ----\n * ----------------------------------------------------------------\n */\n static void\n! ProcessQueryDesc(QueryDesc *queryDesc, Node *limoffset, Node *limcount)\n {\n \tQuery\t *parseTree;\n \tPlan\t *plan;\n***************\n*** 330,336 ****\n \t *\t actually run the plan..\n \t * ----------------\n \t */\n! \tExecutorRun(queryDesc, state, EXEC_RUN, 0);\n \n \t/* save infos for EndCommand */\n \tUpdateCommandInfo(operation, state->es_lastoid, state->es_processed);\n--- 330,336 ----\n \t *\t actually run the plan..\n \t * ----------------\n \t */\n! \tExecutorRun(queryDesc, state, EXEC_RUN, limoffset, limcount);\n \n \t/* save infos for EndCommand */\n \tUpdateCommandInfo(operation, state->es_lastoid, state->es_processed);\n***************\n*** 373,377 ****\n \t\tprint_plan(plan, parsetree);\n \t}\n \telse\n! \t\tProcessQueryDesc(queryDesc);\n }\n--- 373,377 ----\n \t\tprint_plan(plan, parsetree);\n \t}\n \telse\n! \t\tProcessQueryDesc(queryDesc, parsetree->limitOffset, parsetree->limitCount);\n }\ndiff -cr src.orig/include/executor/executor.h src/include/executor/executor.h\n*** src.orig/include/executor/executor.h\tFri Oct 16 11:53:56 1998\n--- src/include/executor/executor.h\tFri Oct 16 12:04:17 1998\n***************\n*** 83,89 ****\n * prototypes from functions in execMain.c\n */\n extern TupleDesc ExecutorStart(QueryDesc *queryDesc, EState *estate);\n! extern TupleTableSlot *ExecutorRun(QueryDesc *queryDesc, EState *estate, int feature, int count);\n extern void ExecutorEnd(QueryDesc *queryDesc, EState *estate);\n extern HeapTuple ExecConstraints(char *caller, Relation rel, HeapTuple tuple);\n #ifdef QUERY_LIMIT\n--- 83,89 ----\n * prototypes from functions in execMain.c\n */\n extern TupleDesc ExecutorStart(QueryDesc *queryDesc, EState *estate);\n! extern TupleTableSlot *ExecutorRun(QueryDesc *queryDesc, EState *estate, int feature, Node *limoffset, Node *limcount);\n extern void ExecutorEnd(QueryDesc *queryDesc, EState *estate);\n extern HeapTuple ExecConstraints(char *caller, Relation rel, HeapTuple tuple);\n #ifdef QUERY_LIMIT\n",
"msg_date": "Thu, 22 Oct 1998 16:50:40 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help (the LIMIT stuff)"
},
{
"msg_contents": "I also would like to use Web-optimized postgresql in production.\nPatched postgres (second trial) passed all regression test on my\nLinux box and seems nothing was broken. How about \n--enable-limit option to configure so people could make a choice ?\n\n\tRegards,\n\n\tOleg\n\nOn Thu, 22 Oct 1998, Jan Wieck wrote:\n\n> \n> > We are close to final, and can easily put it in 6.4.1, which I am sure\n> > we will need, and if we split CVS trees, you'll have lots of minor\n> > versions to pick from. :-)\n> >\n> > Seems like it would be a nice minor release item, but the problem is\n> > that minor releases aren't tested as much as major ones. How confident\n> > are you in the code? What do others thing?\n> \n> I regression tested it, and did additional tests in the\n> SPI/PL area. It works. It only touches the parser and the\n> executor. Rules, planner/optimizer just bypass the values in\n> the parsetree. The parser and the executor are parts of\n> Postgres I feel very familiar with (not so in the optimizer).\n> I trust in the code and would use it in a production\n> environment.\n> \n> It's below.\n> \n> \n> Jan\n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n",
"msg_date": "Thu, 22 Oct 1998 20:29:32 +0400 (MSD)",
"msg_from": "Oleg Bartunov <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help (the LIMIT stuff)"
},
{
"msg_contents": "> I also would like to use Web-optimized postgresql in production.\n> Patched postgres (second trial) passed all regression test on my\n> Linux box and seems nothing was broken. How about \n> --enable-limit option to configure so people could make a choice ?\n> \n> \tRegards,\n\nIt will be in 6.4, or in a minor release soon after 6.4.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 13:30:02 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help (the LIMIT stuff)"
},
{
"msg_contents": "> > Seems like it would be a nice minor release item, but the problem is\n> > that minor releases aren't tested as much as major ones. How confident\n> > are you in the code? What do others thing?\n> \n> I regression tested it, and did additional tests in the\n> SPI/PL area. It works. It only touches the parser and the\n> executor. Rules, planner/optimizer just bypass the values in\n> the parsetree. The parser and the executor are parts of\n> Postgres I feel very familiar with (not so in the optimizer).\n> I trust in the code and would use it in a production\n> environment.\n> \n> It's below.\n\nHaven't forgotten about this. Planned for the first minor release after\n6.4. Hopefully I will have a cnf'ify/OR/palloc fix by then too.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sun, 25 Oct 1998 22:06:52 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql's help (the LIMIT stuff)"
}
] |
[
{
"msg_contents": "Hi all,\n\nI see now \\d shows indexes, I tried to display index structure but it\nshow the wrong\nlength of fields:\n\nTable = figure\n+----------------------------------+----------------------------------+-------+\n| Field | Type |\nLength|\n+----------------------------------+----------------------------------+-------+\n| azienda | char() not null \n| 11 |\n| tipo | char() not null \n| 2 |\n| gruppo | char() not null \n| 2 |\n| inizio_attivita | date not null \n| 4 |\n| fine_attivita | date \n| 4 |\n| codice_figura | char() not null \n| 11 |\n+----------------------------------+----------------------------------+-------+\nIndex: figure_pkey\nhygea=> \\d figure_pkey\n\nTable = figure_pkey\n+----------------------------------+----------------------------------+-------+\n| Field | Type |\nLength|\n+----------------------------------+----------------------------------+-------+\nit should be:\n| azienda | char() \n| -4 |<<-- 11\n| tipo | char() \n| -4 |<<-- 02\n| gruppo | char() \n| -4 |<<-- 02\n| inizio_attivita | date \n| 4 |<<-- 04\n+----------------------------------+----------------------------------+-------+\n\nJose'\n",
"msg_date": "Thu, 22 Oct 1998 10:49:59 +0200",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "psql \\d command"
},
{
"msg_contents": ">\n> Hi all,\n>\n> I see now \\d shows indexes, I tried to display index structure but it\n> show the wrong\n> length of fields:\n>\n> [...]\n>\n> hygea=> \\d figure_pkey\n>\n> Table = figure_pkey\n> +----------------------------------+----------------------------------+-------+\n> | Field | Type |\n> Length|\n> +----------------------------------+----------------------------------+-------+\n> it should be:\n> | azienda | char()\n> | -4 |<<-- 11\n> | tipo | char()\n> | -4 |<<-- 02\n> | gruppo | char()\n> | -4 |<<-- 02\n> | inizio_attivita | date\n> | 4 |<<-- 04\n> +----------------------------------+----------------------------------+-------+\n\n The atttypmod value off all index attributes is 0 in\n pg_attribute. That's the reason why \\d shows this. This\n information is not required for indices because the datum\n given to the index access methods comes from the heap tuples\n and it must already have the correct size if it reaches\n there.\n\n For data types of variable size, an atttypmod value of -1\n means variable, a value of n means size = n-4 (4 is the size\n of the variable length datum header VARHDRSZ). Don't worry,\n you cannot select from an index directly, so it doesn't\n matter. But psql might get enhanced for 6.5 to lookup the\n atttypmod of the indexed field in the table instead of using\n that from the index.\n\n BTW: You might also want to look at the pg_indexes view.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Thu, 22 Oct 1998 12:30:38 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql \\d command"
},
{
"msg_contents": "Hello Jan,\n\ngioved�, 22 ottobre 98, you wrote:\n\n>>\n>> Hi all,\n>>\n>> I see now \\d shows indexes, I tried to display index structure but it\n>> show the wrong\n>> length of fields:\n>>\n>> [...]\n>>\n>> hygea=> \\d figure_pkey\n>>\n>> Table = figure_pkey\n>> +----------------------------------+----------------------------------+-------+\n>> | Field | Type |\n>> Length|\n>> +----------------------------------+----------------------------------+-------+\n>> it should be:\n>> | azienda | char()\n>> | -4 |<<-- 11\n>> | tipo | char()\n>> | -4 |<<-- 02\n>> | gruppo | char()\n>> | -4 |<<-- 02\n>> | inizio_attivita | date\n>> | 4 |<<-- 04\n>> +----------------------------------+----------------------------------+-------+\n\nJW> The atttypmod value off all index attributes is 0 in\nJW> pg_attribute. That's the reason why \\d shows this. This\nJW> information is not required for indices because the datum\nJW> given to the index access methods comes from the heap tuples\nJW> and it must already have the correct size if it reaches\nJW> there.\n\nJW> For data types of variable size, an atttypmod value of -1\nJW> means variable, a value of n means size = n-4 (4 is the size\nJW> of the variable length datum header VARHDRSZ). Don't worry,\nJW> you cannot select from an index directly, so it doesn't\nJW> matter. But psql might get enhanced for 6.5 to lookup the\nJW> atttypmod of the indexed field in the table instead of using\nJW> that from the index.\n\nI'm not worry, only a aesthetical question. Thanks for reply.\n\n\n",
"msg_date": "Thu, 22 Oct 1998 15:11:49 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re[2]: [HACKERS] psql \\d command"
},
{
"msg_contents": "> Hi all,\n> \n> I see now \\d shows indexes, I tried to display index structure but it\n> show the wrong\n> length of fields:\n> \n> Table = figure\n> +----------------------------------+----------------------------------+-------+\n> | Field | Type |\n> Length|\n> +----------------------------------+----------------------------------+-------+\n> | azienda | char() not null \n> | 11 |\n> | tipo | char() not null \n> | 2 |\n> | gruppo | char() not null \n> | 2 |\n> | inizio_attivita | date not null \n> | 4 |\n> | fine_attivita | date \n> | 4 |\n> | codice_figura | char() not null \n> | 11 |\n> +----------------------------------+----------------------------------+-------+\n> Index: figure_pkey\n> hygea=> \\d figure_pkey\n> \n> Table = figure_pkey\n> +----------------------------------+----------------------------------+-------+\n> | Field | Type |\n> Length|\n> +----------------------------------+----------------------------------+-------+\n> it should be:\n> | azienda | char() \n> | -4 |<<-- 11\n> | tipo | char() \n> | -4 |<<-- 02\n> | gruppo | char() \n> | -4 |<<-- 02\n> | inizio_attivita | date \n> | 4 |<<-- 04\n> +----------------------------------+----------------------------------+-------+\n\nAdded to TODO:\n\n\t* psql \\d on index with char()/varchar() fields shows improper length\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Fri, 11 Dec 1998 11:38:39 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql \\d command"
}
] |
[
{
"msg_contents": "Hello, I'm trying to use microseconds in a field time:\nsiranau=> create table toto ( a time );\nCREATE\nsiranau=> insert into toto values ('00:10:45.56');\nINSERT 61641 1\nsiranau=> select * from toto;\na \n--------\n00:10:45\n(1 row)\n\nAs you can see, I cannot get the microseconds value.\n\nSomething more problematic is the following:\n% pg_dump -t toto siranau\n\\connect - siranau\nCREATE TABLE toto (a time);\nCOPY toto FROM stdin;\n00:10:45\n\\.\n\nAs you can see, microseconds values are lost !\n\nPh.R.\n\n-- \n_____________________________________________________________________\nPhilippe Rochat,\t\t\t\tEPFL DI-LBD\nDatabase Laboratory \t\t\t\tCH-1015 LAUSANNE\nSwiss Federal Institute of Technology (EPFL) \ttel:++41 21 693 52 53\nBeep-EPFL: 181 4644\t\t\t\tfax:++41 21 693 51 95\nPrivate: Grammont, 9 1007 LAUSANNE \t\ttel:++41 21 617 03 05 \nmailto:[email protected], http://lbdwww.epfl.ch/~prochat\n",
"msg_date": "Thu, 22 Oct 1998 12:37:35 +0200",
"msg_from": "\"Philippe Rochat (RSR: 318 17 93)\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Time format ? (Really microsecond ??)"
}
] |
[
{
"msg_contents": "How do You do !\n\nI use FreeBSD 3.0 and gmake all your postgresql-2.4 get error\n\nThank you help me !!\n\nMessage in bottom.\n-- \n\n Name: RongGui �ٹ�\n---/\\--------------------------------------------------------------\n--/ \\----- E-mail: [email protected] -----------------------------\n-<Rong>---- Homepage: http://nease.gznet.com/~rong/ ---------------\n--\\ /----- ... Or... http://www.qsl.net/bt2li/ -------------------\n---\\/--------------------------------------------------------------",
"msg_date": "Thu, 22 Oct 1998 06:00:29 -0700",
"msg_from": "XXXX <[email protected]>",
"msg_from_op": true,
"msg_subject": "Help me ,Thank you !!"
}
] |
[
{
"msg_contents": "auth 3a8af432 subscribe pgsql-hackers [email protected]\n",
"msg_date": "Thu, 22 Oct 1998 11:39:51 -0500",
"msg_from": "\"Robert M. Adam, II\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Subscribe verify"
}
] |
[
{
"msg_contents": "Hello,\n\nI am looking for a database system where it is not too\ncomplicate to integrate some special locking mechanisms\nand I am considering Postgres as an candidate.\nI have seen that up to now, Postgres uses relation level\nlocking.\nTo get a rough estimate of how long the integration of\nnew locking methods would take and \nhow complicate this would be, let's take as example the \nintegration of row/page level locking into Postgres.\n\n- which parts of the code would have really to be accessed/changed/modified?\n- what would be the time estimate to do so?\n\nThanks a lot in advance for any answer.\n\nBettina Kemme\n\n-----------------------------------------------------\nBettina Kemme \nInstitute of Information Systems \nETH Zentrum Phone: +41 1 632 7258\nCH-8092 Zuerich Fax: +41 1 632 1172\nSwitzerland E-mail: [email protected]\nhttp://www.inf.ethz.ch/personal/kemme\n\n",
"msg_date": "Thu, 22 Oct 1998 18:54:56 +0200 (MET DST)",
"msg_from": "Bettina Kemme <[email protected]>",
"msg_from_op": true,
"msg_subject": "Locking in Postgres"
}
] |
[
{
"msg_contents": "> I am sorry to bother you with another compliance question (I am sure\n> you get many), but could you direct\n> me to were I can obtain a Y2K compliance statement regarding \n> Postgres95 v6.3.2 for a Sparc/Solaris platform?\n> I need to provide this document to my systems administration team \n> before I can install the dbms. Any help would be greatly appreciated.\n\nHmm. I know what I _think_ about y2k on Postgres (no problem), but I'm\nnot sure what your admin team is looking for here (besides an excuse to\navoid work -- *slap* ooh, that wasn't nice was it?).\n\nI'm not a big fan of holding open-source software to a legally binding\nstatement which could confer liability (\"jeesh, you have the source\ndudes, fix it yourself! And contribute back the solution!\" :)\n\nAnyway, if you have an example of what they are looking for, I'd be\nhappy to look at it and write something up. I can imagine something\nwhich says:\n\n1) at the time of writing, Thomas Lockhart, a member of the loose\nconfederation of the Postgres support team, is not aware of having\nreceived any reports of any problems in the Postgres code base related\nto time transitions around Jan 1, 2000.\n\n2) at the time of writing, in limited testing, which is documented in\nthe included regression tests, there have been no reports of problems\nrelated to time transitions around Jan 1, 2000.\n\n3) at the time of writing, to the best of Thomas' knowledge, the\nassumptions Postgres makes about dates specified with a two-digit year\nare documented in the current User's Guide. (the significant transition\nyear is 1970, not 2000.)\n\n3) any y2k problems in the underlying OS related to obtaining \"the\ncurrent time\" may propagate into apparent y2k problems in Postgres.\n\n*aside* I feel so dirty ;)\n\nWhat do the rest of the developers think about this?\n\n - Tom\n",
"msg_date": "Thu, 22 Oct 1998 17:31:10 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: y2k"
},
{
"msg_contents": "> > I am sorry to bother you with another compliance question (I am sure\n> > you get many), but could you direct\n> > me to were I can obtain a Y2K compliance statement regarding \n> > Postgres95 v6.3.2 for a Sparc/Solaris platform?\n> > I need to provide this document to my systems administration team \n> > before I can install the dbms. Any help would be greatly appreciated.\n> \n> Hmm. I know what I _think_ about y2k on Postgres (no problem), but I'm\n> not sure what your admin team is looking for here (besides an excuse to\n> avoid work -- *slap* ooh, that wasn't nice was it?).\n> \n> I'm not a big fan of holding open-source software to a legally binding\n> statement which could confer liability (\"jeesh, you have the source\n> dudes, fix it yourself! And contribute back the solution!\" :)\n> \n> Anyway, if you have an example of what they are looking for, I'd be\n> happy to look at it and write something up. I can imagine something\n> which says:\n\nI have added a mention to the FAQ on the web site, saying the we are Y2K\ncompliant.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 13:58:58 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: y2k"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> I have added a mention to the FAQ on the web site, saying the we are Y2K\n> compliant.\n\n\nEr, is that such a good idea? I might stick my neck out if I am being paid\nfor it but I don't know that I would want lawyers arguing over exactly\nwhat constitutes \"Y2K compliant.\" Sure, we are but that won't stop\nambulance ch^H^H^H^H^H^H^H^H^H^H^Hlawyers from causing us grief.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Thu, 22 Oct 1998 16:10:06 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: y2k"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > I have added a mention to the FAQ on the web site, saying the we are Y2K\n> > compliant.\n> \n> \n> Er, is that such a good idea? I might stick my neck out if I am being paid\n> for it but I don't know that I would want lawyers arguing over exactly\n> what constitutes \"Y2K compliant.\" Sure, we are but that won't stop\n> ambulance ch^H^H^H^H^H^H^H^H^H^H^Hlawyers from causing us grief.\n\nOur software is supplied as-is. Period.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 16:14:44 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: y2k"
},
{
"msg_contents": "\nOn 22-Oct-98 D'Arcy J.M. Cain wrote:\n> Thus spake Bruce Momjian\n>> I have added a mention to the FAQ on the web site, saying the we are Y2K\n>> compliant.\n> \n> \n> Er, is that such a good idea? I might stick my neck out if I am being paid\n> for it but I don't know that I would want lawyers arguing over exactly\n> what constitutes \"Y2K compliant.\" Sure, we are but that won't stop\n> ambulance ch^H^H^H^H^H^H^H^H^H^H^Hlawyers from causing us grief.\n\nBut doesn't that depend entirely on what the meaning of \"is\" is? :)\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Thu, 22 Oct 1998 16:21:18 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: y2k"
},
{
"msg_contents": "\nWhy should we state any differently then most Open Software products:\n\nhttp://www.sendmail.org/faq/section4.html#4.11\nhttp://www.gnu.org/software/year2000.html\nhttp://samba.gorski.net/samba/sambay2k.html\n\nHow do we store our dates? Same as everyone else...seconds since epoch?\nIf so, then its just our 'external representations' that would risk being\noff, no? And, as someone else pointed out, we display ours as 4-digit.\n\n\n\n \nOn Thu, 22 Oct 1998, Thomas G. Lockhart wrote:\n\n> > I am sorry to bother you with another compliance question (I am sure\n> > you get many), but could you direct\n> > me to were I can obtain a Y2K compliance statement regarding \n> > Postgres95 v6.3.2 for a Sparc/Solaris platform?\n> > I need to provide this document to my systems administration team \n> > before I can install the dbms. Any help would be greatly appreciated.\n> \n> Hmm. I know what I _think_ about y2k on Postgres (no problem), but I'm\n> not sure what your admin team is looking for here (besides an excuse to\n> avoid work -- *slap* ooh, that wasn't nice was it?).\n> \n> I'm not a big fan of holding open-source software to a legally binding\n> statement which could confer liability (\"jeesh, you have the source\n> dudes, fix it yourself! And contribute back the solution!\" :)\n> \n> Anyway, if you have an example of what they are looking for, I'd be\n> happy to look at it and write something up. I can imagine something\n> which says:\n> \n> 1) at the time of writing, Thomas Lockhart, a member of the loose\n> confederation of the Postgres support team, is not aware of having\n> received any reports of any problems in the Postgres code base related\n> to time transitions around Jan 1, 2000.\n> \n> 2) at the time of writing, in limited testing, which is documented in\n> the included regression tests, there have been no reports of problems\n> related to time transitions around Jan 1, 2000.\n\n\tAren't 1 & 2 saying the same thing?\n\n> 3) at the time of writing, to the best of Thomas' knowledge, the\n> assumptions Postgres makes about dates specified with a two-digit year\n> are documented in the current User's Guide. (the significant transition\n> year is 1970, not 2000.)\n\n\tURL reference to this section?\n\n> 3) any y2k problems in the underlying OS related to obtaining \"the\n> current time\" may propagate into apparent y2k problems in Postgres.\n\n\tThis is basically what I read on most of the y2k statements...\n\n> What do the rest of the developers think about this?\n\n\tThe only thing I'd mention/provide is a URL to the section of the\nUser's Guide so that ppl dont' have to go searching for it...other then\nthat, I'd say it sounds both accurate to what we know at this time, while\nnot leaving any of us open to \"but, hey, you said there wouldn't be any\nproblems\"...\n\n\tMight it not be wise to add in a comment dealing with the\nversion(s) of PostgreSQL that this pertains to? Something like\n\"referencing the currently released as well as development source trees\"?\nThe only concern would be someone popping up and mentioning 1.01, cause,\nwell, they are still running that?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Fri, 23 Oct 1998 00:02:17 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: y2k"
},
{
"msg_contents": "> http://www.gnu.org/software/year2000.html\n> > 1) at the time of writing, Thomas Lockhart, a member of the loose\n> > confederation of the Postgres support team, is not aware of having\n> > received any reports of any problems in the Postgres code base \n> > related to time transitions around Jan 1, 2000.\n> > 2) at the time of writing, in limited testing, which is documented \n> > in the included regression tests, there have been no reports of \n> > problems related to time transitions around Jan 1, 2000.\n> Aren't 1 & 2 saying the same thing?\n\nThey probably are, but I didn't mean them to. I wanted (1) to refer to\nmy knowledge of the code base, and it's behavior on my machine. I wanted\n(2) to refer to the results of the regression testing on a wider mix of\nmachines, with better docs on what was actually tested and with (a lack\nof) reports of problems from other users.\n\nIt seemed reasonable to (try to) say what I knew for sure, and put my\nname on it, rather than have the group make a statement. But I like the\nGnu statement, and we should refer to that.\n\nProbably better to make a short statement and then refer to the Gnu site\nfor related info. I'll put it in the docs, after running it by the\nhackers group again.\n\n> > 3) any y2k problems in the underlying OS related to obtaining \"the\n> > current time\" may propagate into apparent y2k problems in Postgres.\n> This is basically what I read on most of the y2k statements...\n> The only thing I'd mention/provide is a URL to the section of the\n> User's Guide so that ppl dont' have to go searching for it...other \n> than that, I'd say it sounds both accurate to what we know at this \n> time, while not leaving any of us open to \"but, hey, you said there \n> wouldn't be any problems\"...\n> \n> Might it not be wise to add in a comment dealing with the\n> version(s) of PostgreSQL that this pertains to? Something like\n> \"referencing the currently released as well as development source \n> trees\"? The only concern would be someone popping up and mentioning \n> 1.01, cause, well, they are still running that?\n\nOK, all sounds good. We'll have something in the v6.4 docs.\n\n - Tom\n",
"msg_date": "Fri, 23 Oct 1998 04:29:58 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: y2k"
},
{
"msg_contents": "On Fri, 23 Oct 1998, Thomas G. Lockhart wrote:\n\n> Probably better to make a short statement and then refer to the Gnu site\n> for related info. I'll put it in the docs, after running it by the\n> hackers group again.\n\n\tActually, there is a really nice document that I found concerning\nthe whole Y2K issue that we also might want to refer to. It basically\nrefers to the \"myths and lies\" concerning Y2K, and I found it to be both\naccurate and eye-opening :)\n\nhttp://language.perl.com/news/y2k.html\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Fri, 23 Oct 1998 02:41:25 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: y2k"
},
{
"msg_contents": "> > I'll put it in the docs, after running it by the\n> > hackers group again.\n\n\nY2K Statement\n\n Author: Written by Thomas Lockhart on 1998-10-22.\n\nThe PostgreSQL Global Development Team provides the Postgres software\ncode tree as a public service, without warranty and without liability\nfor it's behavior or performance.\n\nHowever, at the time of writing: \n\no The author, a volunteer on the Postgres support team since November,\n1996, is not aware of any problems in the Postgres code base related to\ntime transitions around Jan 1, 2000 (Y2K). \n\no The author is not aware of any reports of Y2K problems uncovered in\nregression testing or in other field use of recent or current versions\nof Postgres. We might have expected to hear about problems if they\nexisted, given the installed base and the active participation of users\non the support mailing lists. \n\no To the best of the author' knowledge, the assumptions Postgres makes\nabout dates specified with a two-digit year are documented in the\ncurrent User's Guide in the chapter on data types. For two-digit years,\nthe significant transition year is 1970, not 2000; i.e. \"70-01-01\" is\ninterpreted as \"1970-01-01\", whereas \"69-01-01\" is interpreted as\n\"2069-01-01\".\n\no Any Y2K problems in the underlying OS related to obtaining \"the\ncurrent time\" may propagate into apparent Y2K problems in Postgres. \n\nRefer to The Gnu Project and The Perl Institute for further discussion\nof Y2K issues, particularly as it relates to open source, no fee\nsoftware. \n\n\nNote that in the html docs both references above contain URLs. Comments?\nSuggestions?\n\n - Tom\n",
"msg_date": "Sat, 24 Oct 1998 14:08:52 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: y2k"
}
] |
[
{
"msg_contents": "Hi,\n\n there seems to be a problem with the relation statistics in\n pg_class. Could someone explain why this happens?\n\n doc=> vacuum;\n VACUUM\n doc=> select relname, relpages, reltuples from pg_class\n doc-> where relname = 'doc_wordref';\n relname |relpages|reltuples\n -----------+--------+---------\n doc_wordref| 1099| 136027\n (1 row)\n\n -- ******** That's right\n\n doc=> explain select distinct refpage from doc_wordref\n doc-> where refword ~ '^a';\n NOTICE: QUERY PLAN:\n\n Unique\n -> Sort\n -> Index Scan using doc_wordref_word_idx on doc_wordref\n\n EXPLAIN\n\n -- ******** As expected\n\n doc=> select distinct refpage from doc_wordref\n doc-> where refword ~ '^a';\n refpage\n -------\n 2\n 3\n ...\n (164 rows)\n\n doc=> select relname, relpages, reltuples from pg_class\n doc-> where relname = 'doc_wordref';\n relname |relpages|reltuples\n -----------+--------+---------\n doc_wordref| 0| 0\n (1 row)\n\n -- ******** Ooops - where are they gone?\n\n doc=> explain select distinct refpage from doc_wordref\n dos-> where refword ~ '^a';\n NOTICE: QUERY PLAN:\n\n Unique\n -> Sort\n -> Index Scan using doc_wordref_word_idx on doc_wordref\n\n -- ******** Doesn't matter in the same connection, so reconnect\n\n EXPLAIN\n doc=> \\c -\n connecting to new database: doc\n doc=> explain select distinct refpage from doc_wordref\n dos-> where refword ~ '^a';\n NOTICE: QUERY PLAN:\n\n Unique\n -> Sort\n -> Seq Scan on doc_wordref\n\n -- ******** Boom\n\n EXPLAIN\n doc=>\n\n Why does the SELECT throw away the information about the\n number of pages and tuples from pg_class? Is this a bug or a\n feature? If it's a feature, how can I disable it? It is\n reproduceable.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Thu, 22 Oct 1998 21:33:34 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Bug? relpages, reltuples resets to zero"
},
{
"msg_contents": "> Hi,\n> \n> there seems to be a problem with the relation statistics in\n> pg_class. Could someone explain why this happens?\n> \n> doc=> vacuum;\n> VACUUM\n> doc=> select relname, relpages, reltuples from pg_class\n> doc-> where relname = 'doc_wordref';\n\n> doc=> select relname, relpages, reltuples from pg_class\n> doc-> where relname = 'doc_wordref';\n> relname |relpages|reltuples\n> -----------+--------+---------\n> doc_wordref| 0| 0\n> (1 row)\n> \n> -- ******** Ooops - where are they gone?\n> \n> doc=> explain select distinct refpage from doc_wordref\n> dos-> where refword ~ '^a';\n> NOTICE: QUERY PLAN:\n\nI have seen the optimizer stop using indexes, but could never reproduce\nit, and hoped my mega-patch would have fix it.\n\nMy only guess is that vacuum has changed the buffer cache copy of the\npg_class tuple, but did not mark it as dirty, so it was not written back\nout when removed from the buffer cache. When reloaded after the query,\nthe buffer cache is loaded from the disk copy, and the disk copy has\nzeros, because the vacuum copy was not written to disk.\n\nThe active code is in vacuum.c::vc_updstats:\n\n \n /* XXX -- after write, should invalidate relcache in other backends */\n WriteNoReleaseBuffer(ItemPointerGetBlockNumber(&rtup->t_ctid));\n\n RelationInvalidateHeapTuple(rd, rtup);\n \nThis should be marking the buffer as dirty and written out the buffer to\ndisk, so when it gets reloaded, it has the new vacuum statatistics. It\nis also invalidating the catalog cache, so that doesn't get used for\nstats.\n\nThe code looks fine to me. I can't figure out why that would happen. \nCan you try replacing WriteNoReleaseBuffer() with WriteBuffer() and see\nif that fixes it.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 22 Oct 1998 17:10:47 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Bug? relpages, reltuples resets to zero"
},
{
"msg_contents": "> Hi,\n> \n> there seems to be a problem with the relation statistics in\n> pg_class. Could someone explain why this happens?\n> \n> doc=> vacuum;\n> VACUUM\n> doc=> select relname, relpages, reltuples from pg_class\n> doc-> where relname = 'doc_wordref';\n> relname |relpages|reltuples\n> -----------+--------+---------\n> doc_wordref| 1099| 136027\n> (1 row)\n> \n> -- ******** That's right\n> \n> doc=> explain select distinct refpage from doc_wordref\n> doc-> where refword ~ '^a';\n> NOTICE: QUERY PLAN:\n> \n\nAttached is a patch to vacuum.c I would like you to try. I wonder if\ndirty pages are somehow never getting written out because of vacuum's\nuse of transactions.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n*** ./backend/commands/vacuum.c.orig\tThu Oct 22 17:14:37 1998\n--- ./backend/commands/vacuum.c\tThu Oct 22 17:15:35 1998\n***************\n*** 1875,1886 ****\n \t\theap_close(sd);\n \t}\n \n- \t/* XXX -- after write, should invalidate relcache in other backends */\n- \tWriteNoReleaseBuffer(ItemPointerGetBlockNumber(&rtup->t_ctid));\n- \n \tRelationInvalidateHeapTuple(rd, rtup);\n \n! \tReleaseBuffer(buffer);\n \theap_close(rd);\n }\n \n--- 1875,1885 ----\n \t\theap_close(sd);\n \t}\n \n \tRelationInvalidateHeapTuple(rd, rtup);\n \n! \t/* XXX -- after write, should invalidate relcache in other backends */\n! \tWriteBuffer(ItemPointerGetBlockNumber(&rtup->t_ctid));\n! \n \theap_close(rd);\n }",
"msg_date": "Thu, 22 Oct 1998 17:16:33 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Bug? relpages, reltuples resets to zero"
},
{
"msg_contents": "Bruce Momjian wrote:\n\n> I have seen the optimizer stop using indexes, but could never reproduce\n> it, and hoped my mega-patch would have fix it.\n>\n> My only guess is that vacuum has changed the buffer cache copy of the\n> pg_class tuple, but did not mark it as dirty, so it was not written back\n> out when removed from the buffer cache. When reloaded after the query,\n> the buffer cache is loaded from the disk copy, and the disk copy has\n> zeros, because the vacuum copy was not written to disk.\n>\n> The active code is in vacuum.c::vc_updstats:\n\n Your guess was right, thanks. But your solution does not work\n :-(\n\n I found a way to easily reproduce the error.\n\n Run VACUUM\n Restart postmaster\n -> relpages and reltuples gone\n\n I think the simple way of modifying the tuple in the page\n does not work. I found that catalog/index.c does the same for\n relpages and reltuples in UpdateStats(). Calling\n UpdateStats() after vc_updstats() as a quick hack solved the\n problem.\n\n I'm now cvsup'ing, then I'll modify vacuum.c to do it the\n same way as index.c does it. I don't know if calling\n UpdateStats() instead is really a good idea, because vacuum\n potentially truncates files and UpdateStats() does\n RelationGetNumberOfBlocks() instead of getting it by\n argument. This might be wrong at that time.\n\n Let's see what happens when vacuum does it harder.\n\n\nLater, Jan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Fri, 23 Oct 1998 12:46:01 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Bug? relpages, reltuples resets to zero"
},
{
"msg_contents": "> I think the simple way of modifying the tuple in the page\n> does not work. I found that catalog/index.c does the same for\n> relpages and reltuples in UpdateStats(). Calling\n> UpdateStats() after vc_updstats() as a quick hack solved the\n> problem.\n\n And now I know the way of modifying them in the buffer is the\n only one to succeed. Doing it like index.c with\n heap_replace() irritates vacuum itself.\n\n Sometimes it's good to check returncodes :-).\n WriteNoReleaseBuffer() (as it's name says) takes a buffer\n number as argument, not a disk block number. Thus, it\n returned FALSE sometimes when called to write buffer # 0.\n\n The attdisbursion is also permanently saved on disk now. It\n had the same problem.\n\n Anyway, the fix is below.\n\n Patch is regression tested.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n*** vacuum.c.orig\tFri Oct 23 16:47:21 1998\n--- vacuum.c\tFri Oct 23 16:41:56 1998\n***************\n*** 1792,1799 ****\n--- 1792,1807 ----\n \t\t\t/* overwrite the existing statistics in the tuple */\n \t\t\tif (VacAttrStatsEqValid(stats))\n \t\t\t{\n+ \t\t\t\tBuffer\t\tabuffer;\n \n+ \t\t\t\t/*\n+ \t\t\t\t * We manipulate the heap tuple in the\n+ \t\t\t\t * buffer, so we fetch it to get the\n+ \t\t\t\t * buffer number\n+ \t\t\t\t */\n+ \t\t\t\tatup = heap_fetch(ad, SnapshotNow, &atup->t_ctid, &abuffer);\n \t\t\t\tvc_setpagelock(ad, ItemPointerGetBlockNumber(&atup->t_ctid));\n+ \t\t\t\tattp = (Form_pg_attribute) GETSTRUCT(atup);\n \n \t\t\t\tif (stats->nonnull_cnt + stats->null_cnt == 0 ||\n \t\t\t\t\t(stats->null_cnt <= 1 && stats->best_cnt == 1))\n***************\n*** 1822,1828 ****\n \t\t\t\tif (selratio > 1.0)\n \t\t\t\t\tselratio = 1.0;\n \t\t\t\tattp->attdisbursion = selratio;\n! \t\t\t\tWriteNoReleaseBuffer(ItemPointerGetBlockNumber(&atup->t_ctid));\n \n \t\t\t\t/* DO PG_STATISTIC INSERTS */\n \n--- 1830,1843 ----\n \t\t\t\tif (selratio > 1.0)\n \t\t\t\t\tselratio = 1.0;\n \t\t\t\tattp->attdisbursion = selratio;\n! \n! \t\t\t\t/*\n! \t\t\t\t * Invalidate the cache for the tuple\n! \t\t\t\t * and write the buffer\n! \t\t\t\t */\n! \t\t\t\tRelationInvalidateHeapTuple(ad, atup);\n! \t\t\t\tWriteNoReleaseBuffer(abuffer);\n! \t\t\t\tReleaseBuffer(abuffer);\n \n \t\t\t\t/* DO PG_STATISTIC INSERTS */\n \n***************\n*** 1875,1884 ****\n \t\theap_close(sd);\n \t}\n \n \tRelationInvalidateHeapTuple(rd, rtup);\n \n! \t/* XXX -- after write, should invalidate relcache in other backends */\n! \tWriteBuffer(ItemPointerGetBlockNumber(&rtup->t_ctid));\n \n \theap_close(rd);\n }\n--- 1890,1904 ----\n \t\theap_close(sd);\n \t}\n \n+ \t/*\n+ \t * Invalidate the cached pg_class tuple and\n+ \t * write the buffer\n+ \t */\n \tRelationInvalidateHeapTuple(rd, rtup);\n \n! \tWriteNoReleaseBuffer(buffer);\n! \n! \tReleaseBuffer(buffer);\n \n \theap_close(rd);\n }\n",
"msg_date": "Fri, 23 Oct 1998 17:29:33 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Bug? relpages, reltuples resets to zero"
},
{
"msg_contents": "> > I think the simple way of modifying the tuple in the page\n> > does not work. I found that catalog/index.c does the same for\n> > relpages and reltuples in UpdateStats(). Calling\n> > UpdateStats() after vc_updstats() as a quick hack solved the\n> > problem.\n> \n> And now I know the way of modifying them in the buffer is the\n> only one to succeed. Doing it like index.c with\n> heap_replace() irritates vacuum itself.\n> \n> Sometimes it's good to check returncodes :-).\n> WriteNoReleaseBuffer() (as it's name says) takes a buffer\n> number as argument, not a disk block number. Thus, it\n> returned FALSE sometimes when called to write buffer # 0.\n> \n> The attdisbursion is also permanently saved on disk now. It\n> had the same problem.\n> \n> Anyway, the fix is below.\n> \n> Patch is regression tested.\n\nApplied. Thanks.\n\nI am glad you were able to find the actual cause, rather than calling\nthat UpdateStats function. Vacuum is too strange in the way it modifies\nthings, and calling a general-purpose function that calls all sorts of\nother functions could be a problem.\n\nOf course, you are right. I was passing in a block number, rather than\na buffer number. I checked 6.3.2, and I had it right there, so somehow\nI got confused in the megapatch on that item. I just checked the\nmega-patch, and that is the only place I used ItemPointerGetBlockNumber\nas a replacement for Buffer.\n\nGlad to see you were able to make sense of the new heap_fetch() syntax\nto get the buffer. The old code was very unclear in these areas. Now\nthat it is in the developers FAQ, it should be better.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Fri, 23 Oct 1998 13:02:27 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Bug? relpages, reltuples resets to zero"
},
{
"msg_contents": "Sorry to be nit-picky, but is pg_indexes proper English. Isn't it\npg_indices?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Fri, 23 Oct 1998 19:00:22 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "pg_indexes"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n\n> Sorry to be nit-picky, but is pg_indexes proper English. Isn't it\n> pg_indices?\n\nIn proper English, \"indices\" is plural of the singular noun \"index\".\nIn American, \"indexes\" is also allowed. (In proper English, \"indexes\"\nis a conjugation of the transitive verb \"index\", of course.)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n",
"msg_date": "24 Oct 1998 09:27:19 +0100",
"msg_from": "Tom Ivar Helbekkmo <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_indexes"
},
{
"msg_contents": "Bruce Momjian wrote:\n\n>\n> Sorry to be nit-picky, but is pg_indexes proper English. Isn't it\n> pg_indices?\n\n I thought that too. I named it just pg_indexes because Oracle\n has ALL_INDEXES, DBA_INDEXES and USER_INDEXES. I wonder why\n I'm allways copying those typos when implementing something\n similar to what they have.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 26 Oct 1998 09:14:15 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] pg_indexes"
}
] |
[
{
"msg_contents": "What's the deadline for enhancements to the Python interface to make it \ninto the 6.4 release?\n\n\t-Michael Robinson\n\n",
"msg_date": "Fri, 23 Oct 1998 10:39:08 +0800 (GMT)",
"msg_from": "Michael Robinson <[email protected]>",
"msg_from_op": true,
"msg_subject": "6.4 interfaces deadline"
},
{
"msg_contents": "Umm... we're in feature freeze now... It may be a little late, sorry.\n\nTaral\n\n> -----Original Message-----\n> From: [email protected]\n> [mailto:[email protected]]On Behalf Of Michael Robinson\n> Sent: Thursday, October 22, 1998 9:39 PM\n> To: [email protected]\n> Subject: [HACKERS] 6.4 interfaces deadline\n> \n> \n> What's the deadline for enhancements to the Python interface to make it \n> into the 6.4 release?\n> \n> \t-Michael Robinson\n\n",
"msg_date": "Thu, 22 Oct 1998 21:45:13 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] 6.4 interfaces deadline"
},
{
"msg_contents": "On Fri, 23 Oct 1998, Michael Robinson wrote:\n\n> What's the deadline for enhancements to the Python interface to make it \n> into the 6.4 release?\n\n\t6 weeks ago, sorry :(\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Fri, 23 Oct 1998 00:02:59 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] 6.4 interfaces deadline"
},
{
"msg_contents": "Thus spake Michael Robinson\n> What's the deadline for enhancements to the Python interface to make it \n> into the 6.4 release?\n\nActually, that's a separate project. Once 6.4 is finished I'll be\nstarting to put together the next release of PyGreSQL. If you\nhave enhancements you should send them directly to me.\n\nYou may also want to get on the mailing list. Send email with the\nline \"subscribe pygresql\" to [email protected] to get on (if you\naren't already.)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Fri, 23 Oct 1998 07:26:07 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] 6.4 interfaces deadline"
}
] |
[
{
"msg_contents": "(From John Sharp, trying a Solaris installation of standalone psqlODBC)\n\n> There are significant problems in this psqlodbc installation. Since I\n> already have a successful installation of iodbc-2.12|psqlodbc.0249, I\n> haven't pursued the new installation to a sucessful conclusion yet.\n> \n> The configuation utility correctly picks solaris_sparc_gcc as a\n> target, unlike postgres-6.3.2 (I had to provide that info). However\n> it maps this to the target sparc_solaris not solaris_sparc.\n\nIs this a feature of the main Postgres configuration also? Has anyone\nelse had trouble or success with Sparc Solaris for the main\ndistribution? I'm not aware of anything different in the ODBC stuff.\n\n> I found one compile error in socket.c. This was because this is no\n> definition for INADDR_NONE. This compiled correction when I added\n> #define INADDR_NONE -1\n> To config.h. Note: defining this as \"(inaddr_t) -1\" doesn't work on\n> this gcc.\n> I did not look at the other files very much, except that I would still\n> have to make changes to gpps.c to make it compatable with the way ini\n> files are read in iodbc-2.12.\n\nWhat changes? I'm running the same version of iodbc and am not aware of\nany file incompatibilities, which makes it hard to guess at your\nproblems.\n\n> If you make corrections to this installation, I would be willing to\n> try again. But for now, I'll stay with what is working.\n\nThanks for looking at it John.\n\n - Tom\n",
"msg_date": "Fri, 23 Oct 1998 05:13:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: proposed psqlodbc installation doc"
}
] |
[
{
"msg_contents": "\nHello,\n\nLet me start by congratulating everyone involved in the Postgresql project for a job well done. I have thoroughly researched the bug in 6.3.2 with the ODBC ORDER BY as well as the patch supplied by insight for the backend. With the help of the debug information provided by another poster I have devised a simple (maybe stupid) fix which seems to have fixed all the problems related to destroydb and delete from crashing the backend. The change to the patch file is simple :\n\nInstead of using :\n\n+ \t\ttupType = j->jf_cleanTupType; /* Added by [email protected] 5/20/98 */\n\nI changed the line of the patch file to say ...\n\n+ \t\tif (operation != CMD_DELETE) {tupType = j->jf_cleanTupType;} /* Added by [email protected] 5/20/98 modified by Francois Machabee 10/21/1998 */\n\nand voila the backend stopped crashing and the operations are completed correctly (I checked via psql \\l and also doing a vacuum to check the integrity of the db).\n\nI hope this helps people in the community.\n\nFran�ois Machab�e\nNexus Informatique & Marketing, Inc. / Nexus Information Systems & Marketing, Inc.\nWWW: Http://www.nexus-is.qc.ca\nEmail: mailto:[email protected]\n\n",
"msg_date": "Fri, 23 Oct 1998 07:54:30 -0400",
"msg_from": "Francois Machabee <[email protected]>",
"msg_from_op": true,
"msg_subject": "DestroyDB Patch *** Update"
},
{
"msg_contents": "> Let me start by congratulating everyone involved in the Postgresql \n> project for a job well done. I have thoroughly researched the bug in \n> 6.3.2 with the ODBC ORDER BY as well as the patch supplied by insight \n> for the backend. With the help of the debug information provided by \n> another poster I have devised a simple (maybe stupid) fix which seems \n> to have fixed all the problems related to destroydb and delete from \n> crashing the backend. The change to the patch file is simple :\n\nCould you please send a new patch file? Use \"diff -c\" to generate it...\n\nDavid, have you had a chance to look at this? Is it still a problem?\n\n - Tom\n",
"msg_date": "Fri, 23 Oct 1998 13:50:40 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] DestroyDB Patch *** Update"
}
] |
[
{
"msg_contents": "Hi there,\n\nHere is the proper diff -c patch file for the previously mentionned patch :\n\n*** junkfilter_patch.txt\tFri Oct 23 10:26:14 1998\n--- patch-fm\tThu Oct 22 09:17:34 1998\n***************\n*** 29,35 ****\n \t\tJunkFilter *j = (JunkFilter *) ExecInitJunkFilter(targetList);\n \t\testate->es_junkFilter = j;\n + \n! + \t\ttupType = j->jf_cleanTupType; /* Added by [email protected] 5/20/98 */\n \t}\n \telse\n \t\testate->es_junkFilter = NULL;\n--- 29,35 ----\n \t\tJunkFilter *j = (JunkFilter *) ExecInitJunkFilter(targetList);\n \t\testate->es_junkFilter = j;\n + \n! + \t\tif (operation != CMD_DELETE) {tupType = j->jf_cleanTupType;} /* Added by [email protected] 5/20/98 modified by Francois Machabee 10/21/1998 */\n \t}\n \telse\n \t\testate->es_junkFilter = NULL;\n\n\n\nFran�ois Machab�e\nNexus Informatique & Marketing, Inc. / Nexus Information Systems & Marketing, Inc.\nWWW: Http://www.nexus-is.qc.ca\nEmail: mailto:[email protected]\n\n-----Original Message-----\nFrom:\tThomas G. Lockhart [SMTP:[email protected]]\nSent:\tFriday, October 23, 1998 9:51 AM\nTo:\tFrancois Machabee; David Hartwig\nCc:\t'[email protected]'\nSubject:\tRe: [HACKERS] DestroyDB Patch *** Update\n\n> Let me start by congratulating everyone involved in the Postgresql \n> project for a job well done. I have thoroughly researched the bug in \n> 6.3.2 with the ODBC ORDER BY as well as the patch supplied by insight \n> for the backend. With the help of the debug information provided by \n> another poster I have devised a simple (maybe stupid) fix which seems \n> to have fixed all the problems related to destroydb and delete from \n> crashing the backend. The change to the patch file is simple :\n\nCould you please send a new patch file? Use \"diff -c\" to generate it...\n\nDavid, have you had a chance to look at this? Is it still a problem?\n\n - Tom\n\n",
"msg_date": "Fri, 23 Oct 1998 10:29:46 -0400",
"msg_from": "Francois Machabee <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] DestroyDB Patch *** Update"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> Hi there,\n> \n> Here is the proper diff -c patch file for the previously mentionned patch :\n> \n> *** junkfilter_patch.txt\tFri Oct 23 10:26:14 1998\n> --- patch-fm\tThu Oct 22 09:17:34 1998\n> ***************\n> *** 29,35 ****\n> \t\tJunkFilter *j = (JunkFilter *) ExecInitJunkFilter(targetList);\n> \t\testate->es_junkFilter = j;\n> + \n> ! + \t\ttupType = j->jf_cleanTupType; /* Added by [email protected] 5/20/98 */\n> \t}\n> \telse\n> \t\testate->es_junkFilter = NULL;\n> --- 29,35 ----\n> \t\tJunkFilter *j = (JunkFilter *) ExecInitJunkFilter(targetList);\n> \t\testate->es_junkFilter = j;\n> + \n> ! + \t\tif (operation != CMD_DELETE) {tupType = j->jf_cleanTupType;} /* Added by [email protected] 5/20/98 modified by Francois Machabee 10/21/1998 */\n> \t}\n> \telse\n> \t\testate->es_junkFilter = NULL;\n\nI can't apply this one either. First, it is not a patch, but a patch on\na patch. The line that starts \"! +\" says that. Second, am I supposed\nto guess what file this is supposed to be applied to? I can't figure it\nout, but it could be execMain.c. And third, it does not match any code\nin the executor in the 6.4 beta tree. Please send a patch that can\napply against the 6.4 beta on ftp.postgresql.org.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 26 Oct 1998 22:25:18 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] DestroyDB Patch *** Update"
}
] |
[
{
"msg_contents": "I just finished profiling an application that does a lot of SELECTs from\na Postgres database. I expected to find that parsing and conversion of\nthe data would be a big chunk of its runtime; I was thinking about\nconverting from plain ASCII SELECTs to binary cursor FETCHes, and wanted\nto see what it would buy.\n\nWell, I learned something. Parsing of the data is pretty cheap.\nWhat's chewing up a surprisingly large percentage of the runtime is all\nthe malloc()s and free()s that libpq does while building and releasing\nPGresult objects.\n\nIt looks like we could improve the speed of reading data *significantly*\nif libpq didn't malloc space separately for each and every field value.\nAll those small, odd-sized, short-lived malloc blocks are hard on the\naverage malloc library, or at least on the one I have.\n\nSince PGresult is now an opaque structure, it shouldn't be hard to\nchange things so that it allocates blocks of, say, a few K at a time,\nand then fills those blocks with field values. A little wastage of\nspace shouldn't be a problem, especially since (I suppose) few apps keep\nPGresults around for very long.\n\nI'll probably undertake this improvement sometime soon, but I thought\nI'd mention it for the archives, in case I get run over by a truck or\nsomething before I get to it.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Fri, 23 Oct 1998 20:27:23 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Potential performance improvement in libpq"
}
] |
[
{
"msg_contents": "Attached is the up-to-date 6.4 features list.\n\nI will keep this up-to-date, until the final release.\n\n\n---------------------------------------------------------------------------\n\n\nCHANGES IN THE 6.4 RELEASE\n--------------------------\n\nBug Fixes\n---------\nFix for a tiny memory leak in PQsetdb/PQfinish(Bryan)\nRemove char2-16 data types, use char/varchar(Darren)\nPqfn not handles a NOTICE message(Anders)\nShort-term locks now retry locking at random times(David)\nFix up \"ISO-style\" timespan decoding and encoding(Thomas)\nFix problem with table drop after rollback of transaction(Vadim)\nChange error message and remove non-functional update message(Vadim)\nFix for COPY array checking\nFix for SELECT 1 UNION SELECT NULL\nFix for buffer leaks in large object calls(Pascal)\nChange owner from oid to int4 type(Bruce)\nFix a bug in the oracle compatibility functions btrim() ltrim() and rtrim()\nFix for usernames longer than eight characters(Tom)\nFix for shared invalidation cache overflow(Massimo)\nPrevent file descriptor leaks in failed COPY's(Bruce)\nFix problem with username/passwords over 8 characters(Tom)\n\nEnhancements\n------------\nUpgrade ecpg and ecpglib,see src/interfaces/ecpc/ChangeLog(Michael)\nShow the index used in an explain(Zeugswetter)\nMulti-byte awareness of many data types and functions, via configure(Tatsuo)\nNew configure --with-mb option(Tatsuo)\nNew initdb --pgencoding option(Tatsuo)\nNew createdb -E multibyte option(Tatsuo)\nSelect version(); now returns PostgreSQL version(Jeroen)\nLibpq now allows asynchronous clients(Tom)\nAllow cancel from client of backend query(Tom)\nPsql now cancels query with Control-C(Tom)\nDefine \"tz_hour\" and \"tz_minute\" arguments to date_part()(Thomas)\nAdd routines to convert between varchar and bpchar(Thomas)\nAdd routines to allow sizing of varchar and bpchar into target columns(Thomas)\nAdd bit flags to support timezonehour and minute in data retrieval(Thomas)\nAllow more variations on valid floating point numbers (e.g. \".1\", \"1e6\")(Thomas)\nFixes for unary minus parsing with leading spaces(Thomas)\nImplement TIMEZONE_HOUR, TIMEZONE_MINUTE per SQL92 specs(Thomas)\nCheck for and properly ignore FOREIGN KEY column constraints(Thomas)\nDefine USER as synonym for CURRENT_USER per SQL92 specs(Thomas)\nEnable HAVING clause but no fixes elsewhere yet.\nMake \"char\" type a synonym for \"char(1)\" (actually implemented as bpchar)(Thomas)\nSave string type if specified for DEFAULT clause handling(Thomas)\nCoerce operations involving different data types(Thomas)\nAllow some index use for columns of different types(Thomas)\nAdd capabilities for automatic type conversion(Thomas)\nCleanups for large objects, so file is truncated on open(Peter)\nReadline cleanups(Tom)\nAllow psql \\f \\ to make spaces as delimiter(Bruce)\nPass pg_attribute.atttypmod to the frontend for column field lengths(Tom,Bruce)\nMsql compatibility library in /contrib(Aldrin)\nRemove the requirement that ORDER/GROUP BY clause identifiers be \nincluded in the target list(David)\nConvert columns to match columns in UNION clauses(Thomas)\nRemove fork()/exec() and only do fork()(Bruce)\nJdbc cleanups(Peter)\nShow backend status on ps command line(only works on some platforms)(Bruce)\nPg_hba.conf now has a sameuser option in the database field\nMake lo_unlink take oid param, not int4\nNew DISABLE_COMPLEX_MACRO for compilers that can't handle our macros(Bruce)\nLibpgtcl now gets async notifies from libpq(Tom)\nlibpgtcl cleanups(Tom)\nNew locale patch, see docs/README/locale(Oleg)\nFix for pg_dump so CONSTRAINT and CHECK syntax is correct(ccb)\nNew contrib/lo code for large object orphan removal(Peter)\nNew psql command \"SET CLIENT_ENCODING TO 'encoding'\" for multi-bytes\nfeature, see /doc/README.mb(Tatsuo)\n/contrib/noupdate code to revoke update permission on a column\nLibpq can now be compiled on win32(Magnus)\nAdd PQsetdbLogin() in libpq\nTwo styles we agreed upon for database descriptors(Thomas)\nNew 8-byte integer type, checked by configure for OS support(Thomas)\nBetter support for quited table/column names(Thomas)\nSurround table and column names with double-quotes(Thomas) in generated \n\tsql code to preserve case (SQL92 syntax)(Thomas)\nNew libpqrequestCancel(Tom)\nPQreset() now works with passwords(Tom)\nHandle case of GROUP BY target list column number out of range(David)\nAllow UNION in subselects\nAdd auto-size to screen to \\d? commands(Bruce)\nUse UNION to show all \\d? results in one query(Bruce)\nAdd \\d? field search feature(Bruce)\nPg_dump issues fewer \\connect requests(Tom)\nDocument pg_dump -z flag in manual page(Tom)\nAdd HAVING clause with full support for subselects and unions(Stephan)\nFull text indexing routines in contrib/fulltextindex(Marteen)\nTransaction ids now stored in shared memory(Vadim)\nNew PGCLIENTENCODING when issuing COPY command(Tatsuo)\nSupport for SQL92 syntax \"SET NAMES\"(Tatsuo)\nSupport for LATIN2-5(Tatsuo)\nAdd UNICODE regression test case(Tatsuo)\nLock manager cleanup, new locking modes for LLL(Vadim)\nAllow index use with OR clauses(Bruce)\nAllows \"SELECT NULL ORDER BY 1;\"\nExplain VERBOSE prints the plan, and now pretty-prints the plan to\nthe postmaster log file(Bruce)\nAdd Indices display to \\d command(Bruce)\nAllow GROUP BY on functions(David)\nNew pg_class.relkind for large objects(Bruce)\nNew way to send libpq NOTICE messages to a different location(Tom)\nNew \\w write command to psql(Bruce)\nNew /contrib/findoidjoins scans oid columns to find join relationships(Bruce)\nAllow binary-compatible indices to be considered when checking for valid\nindices for restriction clauses containing a constant(Thomas)\nNew ISBN/ISSN code in /contrib/isbn_issn\nAllow NOT LIKE, IN, NOT IN, BETWEEN, and NOT BETWEEN constraint(Thomas)\nNew rewrite system fixes many problems with rules and views(Jan)\n\t* Rules on relations work\n\t* Event qualifications on insert/update/delete work\n\t* New OLD variable to reference CURRENT, CURRENT will be remove in future\n\t* Update rules can reference NEW and OLD in rule qualifications/actions\n\t* Insert/update/delete rules on views work\n\t* Multiple rule actions are now supported, surrounded by parentheses\n\t* Regular users can create views/rules on tables they have RULE permits\n\t* Rules and views inherit the permissions on the creator\n\t* No rules at the column level\n\t* No UPDATE NEW/OLD rules\n\t* New pg_tables, pg_indexes, pg_rules and pg_views system views\n\t* Only a single action on SELECT rules\n\t* Total rewrite overhaul, perhaps for 6.5\n\t* handle subselects\n\t* handle aggregates on views\n\t* handle insert into select from view works\n\t* \n\nSystem indexes are now multi-key(Bruce)\nOidint2, oidint4, and oidname types are removed(Bruce)\nUse system cache for more system table lookups(Bruce)\nNew backend programming language PL/pgSQL in backend/pl(Jan)\nNew SERIAL data type, auto-creates sequence/index(Thomas)\nEnable assert checking without a recompile(Massimo)\nUser lock enhancements(Massimo)\nNew setval() command to set sequence value(Massimo)\nAuto-remove unix socket file on startup if no postmaster running(Massimo)\nConditional trace package(Massimo)\nNew UNLISTEN command(Massimo)\nPsql and libpq now compile under win32 using win32.mak(Magnus)\nLo_read no longer stores trailing NULL(Bruce)\nIdentifiers are now truncated to 31 characters internally(Bruce)\nCreateuser options now availble on the command line\nPL/pgSQL backend programming language(Jan)\nCode for 64-bit integer supported added, configure tested, int8 type(Thomas)\nPrevent file descriptor leaf from failed COPY(Bruce)\nNew pg_upgrade command(Bruce)\nUpdated /contrib directories(Massimo)\nNew CREATE TABLE DEFAULT VALUES statement available(Thomas)\nNew INSERT INTO TABLE DEFAULT VALUES statement available(Thomas)\nNew DECLARE and FETCH feature(Thomas)\nlibpq's internal structures now not exported(Tom)\nAllow up to 8 key indexes(Bruce)\nRemove ARCHIVE keyword, that is no longer used(Thomas)\npg_dump -n flag to supress quotes around indentifiers\ndisable system columns for views(Jan)\nnet INET and CIDR types for network addresses(TomH, Paul)\nno more double quotes in psql output\npg_dump now dumps views(Terry)\nnew SET QUERY_LIMIT(Tatsuo,Jan)\n\nSource Tree Changes\n-------------------\n/contrib cleanup(Jun)\nInline some small functions called for every row(Bruce)\nAlpha/linux fixes\nHp/UX cleanups(Tom)\nMulti-byte regression tests(Soonmyung.)\nRemove --disabled options from configure\nDefine PGDOC to use POSTGRESDIR by default\nMake regression optional\nRemove extra braces code to pgindent(Bruce)\nAdd bsdi shared library support(Bruce)\nNew --without-CXX support configure option(Brook)\nNew FAQ_CVS\nUpdate backend flowchart in tools/backend(Bruce)\nChange atttypmod from int16 to int32(Bruce, Tom)\nGetrusage() fix for platforms that do not have it(Tom)\nAdd PGUSER to libpq man page\nNS32K platform fixes(Phil Nelson, John Buller)\nSco 7/UnixWare 2.x fixes(Billy,others)\nSparc/Solaris 2.5 fixes(Ryan)\nPgbuiltin.3 is obsolete, move to doc files(Thomas)\nEven more documention(Thomas)\nNextstep support(Jacek)\nAix support(David)\npginterface manual page(Bruce)\nshared libraries all have version numbers\nmerged all OS-specific shared library defines into one file\nsmarter TCL/TK configuration checking(Billy)\nsmarter perl configuration(Brook)\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 24 Oct 1998 00:16:34 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Features list"
},
{
"msg_contents": "> Attached is the up-to-date 6.4 features list.\n\nThis is *impressive*... I think everyone deserves a round of applause...\nFeature improvement lists like this I've seen cause increases in *major*\nversion number :)\n\nThanks a lot :)\n\nTaral\n\n",
"msg_date": "Fri, 23 Oct 1998 23:27:59 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Features list"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > Attached is the up-to-date 6.4 features list.\n> \n> This is *impressive*... I think everyone deserves a round of applause...\n> Feature improvement lists like this I've seen cause increases in *major*\n> version number :)\n> \n> Thanks a lot :)\n\nThis is the usual size for our changes list, ~185 lines. If you look in\nthe HISTORY file, you will see previous releases had a similar list of\nfeatures.\n\nHowever, it does impress me with every release.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 24 Oct 1998 00:43:16 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "> Attached is the up-to-date 6.4 features list.\n> I will keep this up-to-date, until the final release.\n\nHi Bruce. I had converted all of the previous entries (v6.3 and before)\nto sgml to allow inclusion in the Admin Guide. The conversion was really\neasy in that I just laid the entire summary as you have it into a\n<programlisting> markup block.\n\nThe nice thing is that we have a prose summary of the most visible or\nimportant changes just above that, with full markup. If we could do the\nsame for your v6.4 list, then we can include it in the full docs and\nalso have the shorter plain-text version.\n\nTake a look in the new html Admin Guide to see the results.\n\nI'm starting on repackaging the ports and installation info from the\nAdmin Guide into a replacement for INSTALL. Included is an early sample\nof the results for your comments and suggestions (there are a few things\nwhich need to be fixed up). So for v6.4, plain-text INSTALL and HISTORY\nshould be able to be easily generated from sgml sources (I generate RTF\nand then save as formatted text from ApplixWare, roughly the same way I\ndo the Postscript docs).\n\nDo you know if install.sgml and INSTALL are in sync?\n\nbtw, a rough page count shows over 500 pages of docs for this upcoming\nrelease, with some significant new stuff. That's not even counting the\n~100 flat files elsewhere in the source tree :*O\n\n - Tom",
"msg_date": "Sat, 24 Oct 1998 05:40:50 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "On Sat, 24 Oct 1998, Bruce Momjian wrote, in part:\n\n> CHANGES IN THE 6.4 RELEASE\n> --------------------------\n> \n> Bug Fixes\n> ---------\n> \n> Enhancements\n> ------------\n> Enable HAVING clause but no fixes elsewhere yet.\n> Add HAVING clause with full support for subselects and unions(Stephan)\n\nThese two comments raise a question as to whether the having clause\nis fully functional. Is it thought to be fully implemented and bug\nfree?\n\nMarc Zuckman\n\n",
"msg_date": "Sat, 24 Oct 1998 19:34:26 -0400 (EDT)",
"msg_from": "Marc Howard Zuckman <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "> On Sat, 24 Oct 1998, Bruce Momjian wrote, in part:\n> \n> > CHANGES IN THE 6.4 RELEASE\n> > --------------------------\n> > \n> > Bug Fixes\n> > ---------\n> > \n> > Enhancements\n> > ------------\n> > Enable HAVING clause but no fixes elsewhere yet.\n> > Add HAVING clause with full support for subselects and unions(Stephan)\n> \n> These two comments raise a question as to whether the having clause\n> is fully functional. Is it thought to be fully implemented and bug\n> free?\n\nYes. Bug free. It was even available in 6.3.2, but considered buggy.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 24 Oct 1998 20:18:12 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "> On Sat, 24 Oct 1998, Bruce Momjian wrote, in part:\n> \n> > CHANGES IN THE 6.4 RELEASE\n> > --------------------------\n> > \n> > Bug Fixes\n> > ---------\n> > \n> > Enhancements\n> > ------------\n> > Enable HAVING clause but no fixes elsewhere yet.\n> > Add HAVING clause with full support for subselects and unions(Stephan)\n> \n> These two comments raise a question as to whether the having clause\n> is fully functional. Is it thought to be fully implemented and bug\n> free?\n\nFull functional.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 24 Oct 1998 22:41:55 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "Bruce Momjian wrote:\n\n>\n> Attached is the up-to-date 6.4 features list.\n>\n> I will keep this up-to-date, until the final release.\n>\n> Enhancements\n> ------------\n> Upgrade ecpg and ecpglib,see src/interfaces/ecpc/ChangeLog(Michael)\n> Show the index used in an explain(Zeugswetter)\n\n EXPLAIN invokes rule system and shows plan(s) for rewritten\n queries.\n\n> New backend programming language PL/pgSQL in backend/pl(Jan)\n> [...]\n> PL/pgSQL backend programming language(Jan)\n\n No need to mention it twice.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 26 Oct 1998 09:29:41 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "Done.\n\n\n\n> Bruce Momjian wrote:\n> \n> >\n> > Attached is the up-to-date 6.4 features list.\n> >\n> > I will keep this up-to-date, until the final release.\n> >\n> > Enhancements\n> > ------------\n> > Upgrade ecpg and ecpglib,see src/interfaces/ecpc/ChangeLog(Michael)\n> > Show the index used in an explain(Zeugswetter)\n> \n> EXPLAIN invokes rule system and shows plan(s) for rewritten\n> queries.\n> \n> > New backend programming language PL/pgSQL in backend/pl(Jan)\n> > [...]\n> > PL/pgSQL backend programming language(Jan)\n> \n> No need to mention it twice.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 26 Oct 1998 11:48:50 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Features list"
}
] |
[
{
"msg_contents": "I have come across a problem where the code generated by autoconf does not \nsubsitute correct values for the following variables:\n\n\tbindir\t\tsbindir\t\tlibexecdir\n\tdatadir\t\tsysconfdir\tsharedstatedir\n\tlocalstatedir\tlibdir\t\tincludedir\n\toldincludedir\tinfodir\t\tmandir\n\nThe problem is that they reference the variables, 'prefix' and 'exec_prefix' \nbut those variables are not expanded in the subsitutions. For example, \nreferencing 'libdir' in an *.in file will have \"${exec_prefix}/lib\" placed in \nthe resulting output file. It should be \"/usr/local/pgsql/lib\" (if the \ndefault exec_prefix (and prefix) value is used). The first patch \n(uw7-1.patch) will correct the problem, which causes an incorrect \nsrc/pl/plpgsql/src/mklang.sql file to be generated.\n\nThe second patch:\n\n 1. Adds a file (mklang.sql) that can be used to add the language\n definitions for PL/TCL to a database.\n\n 2. Generalizes the mklang.sql in src/pl/plpgsql/test so that it will\n reference the correct location for the plpgsql.so shared library.\n\n 3. Generalizes the test_mklang.sql file in src/pl/tcl/test so that is will\n reference the correct localtion for the pltcl.so shared library.\n\nIMPORTANT: Do not apply the second patch without applying the first patch!\n\n\n\n\n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 |",
"msg_date": "Sat, 24 Oct 1998 03:42:54 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Configure problem (and fix)."
},
{
"msg_contents": " I have come across a problem where the code generated by autoconf does not \n subsitute correct values for the following variables:\n\n\t bindir\t\tsbindir\t\tlibexecdir\n\t datadir\t\tsysconfdir\tsharedstatedir\n\t localstatedir\tlibdir\t\tincludedir\n\t oldincludedir\tinfodir\t\tmandir\n\nThe normal usage is fine for things like Makefiles that know about\nenvironment variables and do their own expansion (i.e., the vast\nmajority of the usage of these variables). In this one instance we\nare constructing *.sql files that can't expand environment variables\nand hence need to do that in advance during configuration.\n\nIt seems much cleaner to do the following:\n\n- keep the unexpanded versions in the Makefiles (since it reduces\n cluttered duplication and uses variables as they should be used for\n indirect references), and\n\n- introduce a new expanded variable for this purpose\n\nThe patches below replace Billie's and accomplish his intent of having\nthe *.sql files refer to the correct libraries without needlessly\nexpanding ALL the variables everywhere.\n\nNote that the only difference from his patches for *.sql files is to\nuse @expanded_libdir@ instead of @libdir@. The difference for\nconfigure.in is to create the variable $expanded_libdir without\naffecting the rest of the \"normal\" variables.\n\nPlease use this set of patches instead of his.\n\nCheers,\nBrook\n\n===========================================================================\n--- configure.in.orig\tFri Oct 23 01:00:15 1998\n+++ configure.in\tSat Oct 24 12:06:13 1998\n@@ -905,6 +905,15 @@\n \tAC_SUBST(USE_ODBC)\n fi\n \n+dnl expand some variables for use in files (e.g., *.sql)\n+dnl without environment variables defined\n+\n+ice_save_exec_prefix=${exec_prefix}\n+test \"x$exec_prefix\" = xNONE && eval exec_prefix=${prefix}\n+eval expanded_libdir=${libdir}\n+exec_prefix=${ice_save_exec_prefix}\n+AC_SUBST(expanded_libdir)\n+\n AC_OUTPUT(\n \tGNUmakefile\n \tMakefile.global\n===========================================================================\n--- pl/plpgsql/src/mklang.sql.in.orig\tSun Oct 11 22:55:03 1998\n+++ pl/plpgsql/src/mklang.sql.in\tSat Oct 24 11:34:16 1998\n@@ -5,7 +5,7 @@\n --\n \n create function plpgsql_call_handler() returns opaque\n-\tas '@libdir@/plpgsql.so'\n+\tas '@expanded_libdir@/plpgsql.so'\n \tlanguage 'C';\n \n create trusted procedural language 'plpgsql'\n===========================================================================\n*** pl/plpgsql/test/mklang.sql.in.orig\tFri Oct 23 23:44:39 1998\n--- pl/plpgsql/test/mklang.sql.in\tFri Oct 23 23:44:00 1998\n***************\n*** 0 ****\n--- 1,14 ----\n+ --\n+ -- PL/pgSQL language declaration\n+ --\n+ -- $Header: /usr/local/cvsroot/pgsql/src/pl/plpgsql/src/mklang.sql.in,v 1.1 1998/10/12 04:55:03 momjian Exp $\n+ --\n+ \n+ create function plpgsql_call_handler() returns opaque\n+ \tas '@expanded_libdir@/plpgsql.so'\n+ \tlanguage 'C';\n+ \n+ create trusted procedural language 'plpgsql'\n+ \thandler plpgsql_call_handler\n+ \tlancompiler 'PL/pgSQL';\n+ \n*** pl/tcl/test/test_mklang.sql.in.orig\tFri Oct 23 23:46:14 1998\n--- pl/tcl/test/test_mklang.sql.in\tSat Oct 24 00:18:35 1998\n***************\n*** 0 ****\n--- 1,9 ----\n+ \n+ create function pltcl_call_handler() returns opaque\n+ \tas '@expanded_libdir@/pltcl.so'\n+ \tlanguage 'C';\n+ \n+ create trusted procedural language 'pltcl'\n+ \thandler pltcl_call_handler\n+ \tlancompiler 'PL/Tcl';\n+ \n*** pl/tcl/mklang.sql.in.orig\tSat Oct 24 00:20:31 1998\n--- pl/tcl/mklang.sql.in\tSat Oct 24 00:18:23 1998\n***************\n*** 0 ****\n--- 1,14 ----\n+ --\n+ -- PL/TCL language declaration\n+ --\n+ -- ## Insert RCS Header Here ##\n+ --\n+ \n+ create function pltcl_call_handler() returns opaque\n+ \tas '@expanded_libdir@/pltcl.so'\n+ \tlanguage 'C';\n+ \n+ create trusted procedural language 'pltcl'\n+ \thandler pltcl_call_handler\n+ \tlancompiler 'PL/Tcl';\n+ \n*** configure.in.orig\tFri Oct 23 23:42:40 1998\n--- configure.in\tFri Oct 23 23:56:13 1998\n***************\n*** 954,958 ****\n--- 954,961 ----\n \tinterfaces/odbc/Makefile.global\n \tpl/plpgsql/src/Makefile\n \tpl/plpgsql/src/mklang.sql\n+ \tpl/plpgsql/test/mklang.sql\n \tpl/tcl/mkMakefile.tcldefs.sh\n+ \tpl/tcl/mklang.sql\n+ \tpl/tcl/test/test_mklang.sql\n )\n",
"msg_date": "Sat, 24 Oct 1998 12:54:03 -0600 (MDT)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem (and fix)."
}
] |
[
{
"msg_contents": "Thus spake Bruce Momjian\n> > I didn't notice anything about primary keys. Not sure if you want\n> > to mention it since it is incomplete but it is an enhancement over\n> > 6.3. I just sent docs to Tom explaining how to utilize it as is.\n> \n> I don't know anything about it. Give me a one-liner.\n\nHmmm. How about this?\n\nFacility (incomplete) for tracking primary keys. To be fully implemented\nin the next release but see [wherever explanation goes] for information\non how to use it now.\n\nNot a one liner though. Hmmm...\n\nCatalogue support for primary keys.\n\nMaybe?\n\nPerhaps others can help. I'll copy the list on this message.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Sat, 24 Oct 1998 11:36:32 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > > I didn't notice anything about primary keys. Not sure if you want\n> > > to mention it since it is incomplete but it is an enhancement over\n> > > 6.3. I just sent docs to Tom explaining how to utilize it as is.\n> > \n> > I don't know anything about it. Give me a one-liner.\n> \n> Hmmm. How about this?\n> \n> Facility (incomplete) for tracking primary keys. To be fully implemented\n> in the next release but see [wherever explanation goes] for information\n> on how to use it now.\n> \n> Not a one liner though. Hmmm...\n> \n> Catalogue support for primary keys.\n> \n> Maybe?\n> \n> Perhaps others can help. I'll copy the list on this message.\n\nWhen someone uses PRIMARY in table creation, why don't we just plop\nsomething in the field?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 24 Oct 1998 11:59:05 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> When someone uses PRIMARY in table creation, why don't we just plop\n> something in the field?\n\nWell, that would be the \"completion\" of the facility. Also, we should\nfill in the proper values in the system catalogues. The latter is probably\neasy but the former rewuires mucking with the table creation code. I don't\nthink it is a lot of work but perhaps not something to start now.\n\nIn any case, we're past the freeze, right?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Sat, 24 Oct 1998 12:48:29 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "> Thus spake Bruce Momjian\n> > When someone uses PRIMARY in table creation, why don't we just plop\n> > something in the field?\n> \n> Well, that would be the \"completion\" of the facility. Also, we should\n> fill in the proper values in the system catalogues. The latter is probably\n> easy but the former rewuires mucking with the table creation code. I don't\n> think it is a lot of work but perhaps not something to start now.\n> \n> In any case, we're past the freeze, right?\n\nI hesitate to add a mention that we have a field for primary key, while\nwe do nothing with it.\n\nThis was brought up, but post-freeze, and since it was totally new,\ncould not be added.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 24 Oct 1998 19:46:16 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": ">\n> > Thus spake Bruce Momjian\n> > > When someone uses PRIMARY in table creation, why don't we just plop\n> > > something in the field?\n> >\n> > Well, that would be the \"completion\" of the facility. Also, we should\n> > fill in the proper values in the system catalogues. The latter is probably\n> > easy but the former rewuires mucking with the table creation code. I don't\n> > think it is a lot of work but perhaps not something to start now.\n> >\n> > In any case, we're past the freeze, right?\n>\n> I hesitate to add a mention that we have a field for primary key, while\n> we do nothing with it.\n>\n> This was brought up, but post-freeze, and since it was totally new,\n> could not be added.\n\n I've seen the relfkeys too and have something in mind for\n FOREIGN KEY for 6.5. 6.5 will be the comeback of attribute\n rules and I think that foreign key checks and the like\n (including ON DELETE CASCADE) could get implemented using\n rules.\n\n Let's pick it up after 6.4 is out.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 26 Oct 1998 10:53:01 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Features list"
},
{
"msg_contents": "> I've seen the relfkeys too and have something in mind for\n> FOREIGN KEY for 6.5. 6.5 will be the comeback of attribute\n> rules and I think that foreign key checks and the like\n> (including ON DELETE CASCADE) could get implemented using\n> rules.\n> \n> Let's pick it up after 6.4 is out.\n\nThat is a temping way to do foreign keys. Nice.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 26 Oct 1998 11:49:41 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Features list"
}
] |
[
{
"msg_contents": "Hi\n\nI posted this message a few days ago on pgsql-admin, alas until now i\nhave not got an answer.\nI cheked the Logs .. and tried to make as mutch sense out of them as\npossible..\n\nA few errors where easy to figure out. a few others are harder.\n\nI guess the illegal value gets put in the DB , with a warning message,\nbut is striped .. and the data is trunched. well i added my previos\npost and my diff file.\n\nRegards\n\nGavin Woodhatch\nSpace4u.com\n\n<Begin Old Msg>\n\n\nHello,\n\nI have included all failed tests , and added my comment ( if for e.g. it\n\nfailed because of some stupid difference)\n\nbut what i dont know is , how bad these errors are , as i have a very\nlimited knowlege on RDBMS.\n\nI am Running Linux Kernel 2.0.35 , PostgreSQL 6.3.2, GLIBC2 version\n2.0.7-19,\ncompiled with egcs 1.1b. on an DEC Alpha\n\ni had to do a \"configure -with-template=linuxalpha\" for postgres to\nwork.\nwhen compiling with the linux template i count eaven do an initdb ..\n(IMHO this would be one for the FAQ)\n\nint2 .. failed ( Two Differnt Error Messages)\n\nint4 .. failed ( see attached .gz file)\nwhat concerns me here is that a value of -727379968 never got put in to\n\nthe DB .. strange\n\noidint2 .. failed ( Two Differnt Error Messages)\n\noidint4 .. failed\ni have the feeling that some invalid data gets put in the table ..\nthere are mor rows than there should be ..\n\nfloat4 .. failed\nsame prob here\n\nfloat8 .. failed\nsame prob here\n\nnumerology .. failed\ngeometry .. failed i guess the same here :(\n\ndatetime .. failed\nreltime .. failed\nabstime .. failed\ntinterval .. failed\nhorology .. failed\n\nsome problem with dates on Linux Alpha i get years +2000 (hey can my\nalpha do time leeps ?)\n\nrandom .. failed\nmisc .. failed\nhmm can't figure out what went wrong here\n\nIf anyone can tell me how bad this stuff is .. or what maybe when wrong\nwhen compiling.\nwould be glad to know.\n\nRegards\n\nGavin Woodhatch\nSpace4u.com",
"msg_date": "Sat, 24 Oct 1998 20:40:31 +0200",
"msg_from": "Gavin <[email protected]>",
"msg_from_op": true,
"msg_subject": "Regression Tests fail "
}
] |
[
{
"msg_contents": "I wanted to preserve table ownerships and access permissions in my\ndatabase, so I tried using pg_upgrade with a script made by\n\"pg_dumpall -s -z\", rather than just \"pg_dumpall -s\" as suggested\nin the man page.\n\nDidn't work. It *would've* worked, except that pg_dumpall tries to\nreload the pg_shadow table via \"COPY FROM stdin\" ... and pg_upgrade very\ncarefully removes that command from the dump script before executing it.\nSo the database still has only the postgres superuser as an allowed\nuser, and psql's attempts to connect as other users fail.\n\nEven if I weren't trying to preserve table ownerships, I would consider\nit a bug in pg_upgrade that the contents of pg_shadow are lost.\n\nI think probably the right answer is for pg_upgrade just to execute\nthe given script, not try to alter it in any way. If you want to defend\nagainst people feeding full dumps rather than -s dumps to pg_upgrade,\nmay I suggest that we just grep the script to see whether it contains\nany copy commands other than the one directed to pg_shadow?\n\nI didn't apply such a fix myself, since someone else might have a better\nidea. I did patch pg_upgrade to notice if the psql command fails,\nhowever. Plowing ahead with modifying the data directories after a\nscript failure is definitely a bad idea. (Good thing I made a regular\ndump file before I tried this ;-))\n\nBTW, I'd suggest that once this glitch is fixed, the man page for\npg_upgrade should recommend -s -z not just -s as the standard flags\nto pg_dumpall.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sat, 24 Oct 1998 16:32:54 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "pg_upgrade bug report"
},
{
"msg_contents": "> I wanted to preserve table ownerships and access permissions in my\n> database, so I tried using pg_upgrade with a script made by\n> \"pg_dumpall -s -z\", rather than just \"pg_dumpall -s\" as suggested\n> in the man page.\n> \n> Didn't work. It *would've* worked, except that pg_dumpall tries to\n> reload the pg_shadow table via \"COPY FROM stdin\" ... and pg_upgrade very\n> carefully removes that command from the dump script before executing it.\n> So the database still has only the postgres superuser as an allowed\n> user, and psql's attempts to connect as other users fail.\n> \n> Even if I weren't trying to preserve table ownerships, I would consider\n> it a bug in pg_upgrade that the contents of pg_shadow are lost.\n> \n> I think probably the right answer is for pg_upgrade just to execute\n> the given script, not try to alter it in any way. If you want to defend\n> against people feeding full dumps rather than -s dumps to pg_upgrade,\n> may I suggest that we just grep the script to see whether it contains\n> any copy commands other than the one directed to pg_shadow?\n> \n> I didn't apply such a fix myself, since someone else might have a better\n> idea. I did patch pg_upgrade to notice if the psql command fails,\n> however. Plowing ahead with modifying the data directories after a\n> script failure is definitely a bad idea. (Good thing I made a regular\n> dump file before I tried this ;-))\n> \n> BTW, I'd suggest that once this glitch is fixed, the man page for\n> pg_upgrade should recommend -s -z not just -s as the standard flags\n> to pg_dumpall.\n> \n\nGood analysis. I have fixed the awk code to skip the COPY for\npg_shadow. Also added the -z mention in the docs.\n\nSo, I guess pg_upgrade works?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 24 Oct 1998 22:47:47 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_upgrade bug report"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> So, I guess pg_upgrade works?\n\nFWIW, I tried it again after you fixed the -z problem, and it seemed to\nwork this time. I haven't beat hard on the resulting database yet...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 14:53:37 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] pg_upgrade bug report "
}
] |
[
{
"msg_contents": "Brook Milligan wrote:\n> \n> The normal usage is fine for things like Makefiles that know about\n> environment variables and do their own expansion (i.e., the vast\n> majority of the usage of these variables). In this one instance we\n> are constructing *.sql files that can't expand environment variables\n> and hence need to do that in advance during configuration.\n> \n> It seems much cleaner to do the following:\n> \n> - keep the unexpanded versions in the Makefiles (since it reduces\n> cluttered duplication and uses variables as they should be used for\n> indirect references), and\n> \n> - introduce a new expanded variable for this purpose\n\nIs it cleaner? It makes sense to me that if I want to use 'libdir' or any of \nthe other defined variables, it should have a value that can be anywhere it is \nneeded without having to depend on the expansion of variables in the place \nwhere it is used. Configure knows the information needed to expand the \nvariables and should do so before making subsitutions. If this is not done, \nit leave the door open for problems to occur. For instance, if after running \nconfigure, the global makefile is modified so that 'prefix' is changed (an \nassumption can be made that since the definitions have unexpanded variables, \nthen this can work without having to re-run configure), then the location of \n'libdir' will be differenct from 'expanded_libdir' and the *.sql files will no \nlonger be correct. The assumption that a variable, such as 'prefix', can be \nchanged in the global makefile (an assumption encouraged by the way configure \nworks without my patch) and still have the system build correctly is incorrect.\n\n> \n> The patches below replace Billie's and accomplish his intent of having\nBTW: That's Billy's ----------^\n\nMy intent is to be able to use the subsitution feature of configure any where \nit is needed without haveing to depend on expansion of embeded variable \noccuring where they are used.\n\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Sat, 24 Oct 1998 20:32:33 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Configure problem (and fix)."
}
] |
[
{
"msg_contents": "I guess it's too late for 6.4 so hopefully there is already a way to\nfind out whether a field was declared as a serial type. Am I stuck\nusing some sort of index name hack?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Sun, 25 Oct 1998 08:50:36 -0500 (EST)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "How do we find serial types"
},
{
"msg_contents": "> I guess it's too late for 6.4 so hopefully there is already a way to\n> find out whether a field was declared as a serial type. Am I stuck\n> using some sort of index name hack?\n\nProbably. I did the serial type implementation as a hash of hooks I\nalready had in the backend (and at the behest of others, so don't blame\nme too much :). It isn't integrated as well as it could/should be, and\nwe should clean this up for v6.5.\n\nIt actually might be related to what we will need to do for primary and\nforeign keys, so might get fixed with that stuff. Jan has indicated an\ninterest in pursuing some of this through the rules system.\n\n - Tom\n",
"msg_date": "Mon, 26 Oct 1998 14:55:44 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] How do we find serial types"
},
{
"msg_contents": "Thus spake Thomas G. Lockhart\n> > I guess it's too late for 6.4 so hopefully there is already a way to\n> > find out whether a field was declared as a serial type. Am I stuck\n> > using some sort of index name hack?\n> \n> Probably. I did the serial type implementation as a hash of hooks I\n> already had in the backend (and at the behest of others, so don't blame\n> me too much :). It isn't integrated as well as it could/should be, and\n> we should clean this up for v6.5.\n\nThat would be good. Mostly I am interested in knowing that it was\ndeclared serial in order to write generic forms handlers and such.\nBasically I want to not insert values into serial types so that the\nsystem will generate the sequential number.\n\nActually, there is another way to do this. If the system were to always\ngenerate a sequential number on insert and ignore any value specified, that\nwould work to. Unfortunately that wouldn't work on a dump and restore.\n\n> It actually might be related to what we will need to do for primary and\n> foreign keys, so might get fixed with that stuff. Jan has indicated an\n> interest in pursuing some of this through the rules system.\n\nAt least with the primary key there is a field so we can kluge it. I\nwrote some documentation that you should now have explaining how to do that.\n\nArgh! I fscked up! I sent the docs to Tom Ivar Helbekkmo by mistake\nbecause he sent me the inet skeleton docs. He explained the error but\nI misread his message. I thought he was sending it on to you but he\nsaid for me to send it on. I'll get that to you right away. Sorry\nabout that. I'll bounce you those messages right away. Did you get the\ninet/cidr regression test queries? I'll bounce that message to you too\njust in case.\n\nAnd I was so happy that I got everything in before the Friday date.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Mon, 26 Oct 1998 11:15:27 -0500 (EST)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] How do we find serial types"
},
{
"msg_contents": "> Argh! I fscked up! I sent the docs to Tom Ivar Helbekkmo by mistake\n> because he sent me the inet skeleton docs. He explained the error but\n> I misread his message. I thought he was sending it on to you but he\n> said for me to send it on. I'll get that to you right away. Sorry\n> about that. I'll bounce you those messages right away. Did you get \n> the inet/cidr regression test queries? I'll bounce that message to \n> you too just in case.\n\nUh, sorry but I haven't seen anything. Still waiting for any/all inet\ndocs and regression tests, and will freeze with or without in one more\nday. But it sounds like you have something, so be sure to send it soon.\n\n> And I was so happy that I got everything in before the Friday date.\n\nHappiness sure is fleeting sometimes, eh? :)\n\n - Tom^H^H^HThomas\n",
"msg_date": "Mon, 26 Oct 1998 16:48:47 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] How do we find serial types"
},
{
"msg_contents": "Thomas G. Lockhart wrote:\n\n>\n> > I guess it's too late for 6.4 so hopefully there is already a way to\n> > find out whether a field was declared as a serial type. Am I stuck\n> > using some sort of index name hack?\n>\n> Probably. I did the serial type implementation as a hash of hooks I\n> already had in the backend (and at the behest of others, so don't blame\n> me too much :). It isn't integrated as well as it could/should be, and\n> we should clean this up for v6.5.\n>\n> It actually might be related to what we will need to do for primary and\n> foreign keys, so might get fixed with that stuff. Jan has indicated an\n> interest in pursuing some of this through the rules system.\n\n I'm not sure if the rule system can address serial types as I\n don't know what is required for them. Should a serial type be\n forced to use nextval('serial') on INSERT and then be\n forbidden to be touched until DELETE?\n\n That's not possible with the current rule system. But the\n force/check could be implemented similar to rewrite rules.\n After parsing we could replace or add the TLE expression in\n INSERT to be a nextval('serial') call if it isn't already\n that. And on UPDATE we could simply throw an ERROR if the\n serial type column appears in the targetlist and isn't a Var\n reference to it's own OLD value. Other behaviour is also\n possible. Since I'm very familiar with screwing around in\n the targetlist from the work in the rule system I would help\n doing so :-)\n\n What I'm planning on the rule system is to bring back\n attribute checks. I think it would be good to have an array\n of up to 8 attribute numbers instead of the one attno\n currently present. Then the actual rule will only be fired,\n if one of the named attributes appears in the targetlist and\n isn't a reference to OLD.same_attr. Having a rule ON UPDATE\n will then only add the rule actions, if one of the attributes\n could change at all (attributes not in the targetlist will\n later be added by the optimizer as Var refs to the OLD tuple\n and thus could never change on this query).\n\n The next thing required then would be the implementation of a\n RAISE statement. This is mostly the same thing as a SELECT,\n but it will use the final result rows to produce elog()\n messages. Trivial to add.\n\n After these enhancements, constraints could be implemented by\n adding special rules built from the CONSTRAINT clauses at\n CREATE TABLE time. The following event/action pairs could\n easily be implemented with the rule system as it already is.\n The attribute checks would only suppress stupid-in-fact-noop\n queries to get generated.\n\n Event: INSERT to a table\n Action: RAISE an ERROR if column values are outside of a\n specified range (fixed list of possible values or a\n subselect from a key table).\n\n Event: UPDATE to a table\n Action: RAISE an ERROR if new colum values are outside of a\n specified range (fixed list of possible values or a\n subselect from a key table).\n\n Event: DELETE from a referenced key table\n Action: Constraint delete rows in other tables that reference\n the deleted key.\n\n OR\n\n RAISE an ERROR telling that the row cannot be deleted\n while it is referenced.\n\n Event: UPDATE on a referenced key table where the values of\n the key fields change\n Action: Constraint delete rows in other tables that reference\n the old key.\n\n OR\n\n RAISE an ERROR telling that the key value cannot be\n changed while it is referenced.\n\n Isn't that all (or already more) we need for FOREIGN KEY and\n ON DELETE CASCADE?\n\n Even if it is possible with the rule system, to let\n references to a key table follow on updates, such a scenario\n isn't easy to setup during CREATE TABLE from the CONSTRAINT\n clauses. This must be setup manually if it is really\n required.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 26 Oct 1998 18:35:58 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] How do we find serial types"
},
{
"msg_contents": "D'Arcy J.M. Cain wrote:\n\n> Actually, there is another way to do this. If the system were to always\n> generate a sequential number on insert and ignore any value specified, that\n> would work to. Unfortunately that wouldn't work on a dump and restore.\n\n The way I described it, using a kind of pre-rewriting, would\n work for dump/restore. Data moved in by COPY doesn't invoke\n any rule since there is no query to rewrite (except you dump\n it as INSERT statements). COPY is a utility statement and\n they aren't rewritten at all. On the other hand we could\n change pg_dump to omit SERIAL and CONSTRAINT information at\n CREATE TABLE and later turn all this on like triggers/rules\n (using ALTER TABLE?) when the data is put back by\n COPY/INSERT.\n\n I think for 6.5 it would also be good to restrict the use of\n COPY to tables the user has RULE permissions for. I think\n such a check doesn't exist already and if I'm right, COPY is\n a way to get around rules ON INSERT.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 26 Oct 1998 18:50:18 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] How do we find serial types"
}
] |
[
{
"msg_contents": "> My stuff is all in, as of yesterday. (BTW, I assume it's sufficient \n> to fix install.sgml, and that the text INSTALL file will be built from \n> that?)\n\nYes. And I'm going through install.sgml at this moment to freshen it up.\nI'll commit changes in the next few minutes, and it would be great if\nothers (Tom and Bruce?) could take a look at it.\n\nI'll try posting a new html version of things on the Postgres web site.\n\n - Tom\n",
"msg_date": "Sun, 25 Oct 1998 16:38:25 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call?"
},
{
"msg_contents": "> I'll try posting a new html version of things on the Postgres web \n> site.\n\n... which I've now done for the admin guide. This contains a chapter on\ninstallation which is from the same source as the text-only version to\nbe put in INSTALL.\n\nCould folks look it over and in particular make sure that it is an\nacceptable substitute for what is now in INSTALL?\n\n - Tom\n",
"msg_date": "Sun, 25 Oct 1998 17:31:17 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call?"
},
{
"msg_contents": "> > My stuff is all in, as of yesterday. (BTW, I assume it's sufficient \n> > to fix install.sgml, and that the text INSTALL file will be built from \n> > that?)\n> \n> Yes. And I'm going through install.sgml at this moment to freshen it up.\n> I'll commit changes in the next few minutes, and it would be great if\n> others (Tom and Bruce?) could take a look at it.\n> \n> I'll try posting a new html version of things on the Postgres web site.\n> \n> - Tom\n\nAs far as I know, the INSTALL and sgml INSTALL are in sync. I have made\nchanges both places, and I don't think others have made changes.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sun, 25 Oct 1998 12:32:31 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Last call?"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> ... which I've now done for the admin guide. This contains a chapter on\n> installation which is from the same source as the text-only version to\n> be put in INSTALL.\n> Could folks look it over and in particular make sure that it is an\n> acceptable substitute for what is now in INSTALL?\n\nA couple minor comments:\n\n1. I think the disk space estimates are now too low. Yesterday I\nmeasured the required space to build/install/do regression test on an\nHPUX 10.20 box, which is probably not too unrepresentative of RISC-type\nworkstations. The source tree got up to 29600K, the installed code +\ninitdb took about 4600K, and the regression test database occupied\n18500K. Furthermore, I did not build Tcl, Perl, or ODBC libraries,\nwhich would certainly have increased the build space and installed code\nsize by a few more megs. Still more: the source tree included the\nformatted doc files (those tar.gz files in doc/) that are currently in\nthe CVS server. I assume that 6.4's formatted doc files will be\nsubstantially bulkier, thanks to Thomas' tireless efforts ;-). And\nstill more: I didn't do an install of the documentation files. So you\nneed to add whatever the installed footprint of the html, ps, and\nman-page files are.\n\nThe estimates given in the INSTALL docs (45MB peak usage during build &\ninstall, 3MB permanent footprint) are probably at least 50% too low.\nWe've been busy little builders, evidently :-).\n\nIt's probably not necessary to repeat the disk space estimates twice,\neither. Step 3a of the install procedure ought to be merged into the\n\"Requirements to run Postgres\" section.\n\n2. In install step 2 (\"Create the Postgres superuser account...\")\nit wouldn't hurt to add \"The owner of the Postgres files can be any\nunprivileged user account. It MUST NOT be root, bin, or any other\naccount with special access rights, as that would create a security risk.\"\nThis point is made later, but the time that an installer needs to know\nit is right here.\n\n3. In step 6 (dumping your old database) I think that the dumpall\ncommand should include -z:\n\t$ src/bin/pg_dump/pg_dumpall -z > db.out\nOtherwise table ownership will not be preserved, which could ruin\nsomeone's day. (BTW, is there really any value in extracting the\npg_dumpall script from the distribution as shown? Most of the problems\nwe've fixed in database dumping have been inside pg_dump, methinks.\nSo it's not clear that this contortion helps anything.)\n\n4. (I mentioned this one already.) Step 19a (start postmaster,\nas a substep of running the regression tests) probably ought to be\na top-level step between 18 and 19. You want people to do it even\nif they skip the regression test.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 26 Oct 1998 19:55:53 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Last call? "
},
{
"msg_contents": "\nOn 27-Oct-98 Tom Lane wrote:\n> \"Thomas G. Lockhart\" <[email protected]> writes:\n>\n> The estimates given in the INSTALL docs (45MB peak usage during build &\n> install, 3MB permanent footprint) are probably at least 50% too low.\n> We've been busy little builders, evidently :-).\n\nOn FreeBSD 2.2.6, I'm just a few K under 50MB for build and install. If\nthe regression tests went beyond that I dunno.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Mon, 26 Oct 1998 20:19:52 -0500 (EST)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Last call?"
},
{
"msg_contents": "> > Could folks look it over and in particular make sure that it is an\n> > acceptable substitute for what is now in INSTALL?\n\nAll good suggestions. Will make some mods.\n\nbtw, I just realized that \"config.sgml\" was not being built into the\nAdmin Guide. It includes sections on enabling locale and on configuring\nKerberos. I've added a section which is essentially the \"configure\noptions\" as is in the installation procedure; I'd propose expanding on\nthat and dropping most of the verbiage in that step of the installation.\n\nI will also add a section on \"make options\", inspired by that same kind\nof info in the ODBC installation docs. It will likely not be complete,\nbut it's a start.\n\nWill try to knock this off soon (tonight?) and check in sources and post\nhtml on the web site. Also, will try another cut at globbing sections\ntogether to make a new INSTALL. Bruce, can you give some feedback on\nthat? I would like to know what info should be added or removed and what\nyou find most annoying about the formatting. I can make adjustments, but\nneed to hear what you would like.\n\nTIA\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 06:43:13 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call?"
}
] |
[
{
"msg_contents": "> I think you messed up a change I had made: the version of install.sgml\n> I committed yesterday had the steps:\n> ...\n> initdb\n> check permissions\n> start postmaster\n> run regression test\n> <substeps>\n> arrange for postmaster autostart\n> ...\n> You seem to have undone that in a way that's broken...\n\nOK, I know I messed it up. I was confused the whole way, and *it's\nbecause you have the same *(&^% username as I've had for 20 years* :)\n\nI had seen that you had committed changes, and then I'd had cvs merging\ndifficulties because I was also making changes to that file. But then\nwhen trying to resolve the patches I chose the wrong block of changes in\nat least one case, and got confused when seeing that \"tgl\" had made the\nchanges.\n\nAnyway, will go through it again.\n\nYour tags were just fine afaik. I have since gone through and put lots\nmore markup in because it was never done the first time when I copied\nINSTALL to become install.sgml.\n\n - Tom^H^H^HThomas\n",
"msg_date": "Sun, 25 Oct 1998 22:53:43 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call?"
}
] |
[
{
"msg_contents": "Hi All,\n\nLooks fine here, although we did change from BST to GMT last night!!\n\nIn general PostgreSQL on S/Linux is looking good.\n\nI'm running kernel 2.0.35 on a mainly Redhat 4.2 system.\n\nThe only regression failures are float8 and geometry, due\nto the usual FP rounding/precision differences.\n\nThanks,\nKeith.\n\n\n\"Billy G. Allie\" <[email protected]>\n> \n> \"Thomas A. Szybist\" wrote:\n> > \n> > I just gave today's (Oct 25) snapshot a try on Sparc Linux. \n> > Looks good except datetime. I'm getting failures due to this type \n> > of thing:\n> > \n> > regression=> SELECT ('today'::datetime );\n> > ?column?\n> > ----------------------------\n> > Sun Oct 25 00:00:00 1998 EDT\n> > (1 row)\n> > \n> > regression=> SELECT ('tomorrow'::datetime - '1 day'::timespan);\n> > ?column?\n> > ----------------------------\n> > Sun Oct 25 01:00:00 1998 EDT\n> > (1 row) \n> > \n> > \n> > I *think* this may because we're not too far into EST yet.\n> > Sound good?\n> > \n> \n> This is not a failure. The date 24 hours (1 day) before 'Mon Oct 26 00:00:00 \n1998 EST' is 'Sun Oct 25 01:00:00 1998 EDT'. You would think it should be 'Sun \nOct 25 00:00:00 1998 EST', but thanks to Daylight Savings Time that datetime \ndoes not exist :-)\n\n",
"msg_date": "Sun, 25 Oct 1998 23:25:14 +0000 (GMT)",
"msg_from": "Keith Parks <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call? "
}
] |
[
{
"msg_contents": "Anyone else notice that the datetime regression test is barfing today?\nFor example,\n\nSELECT ('today'::datetime = ('tomorrow'::datetime - '1 day'::timespan));\n\nis returning FALSE instead of the expected TRUE. I thought I'd broken\nsomething in my installation, but further investigation tells the tale:\n\nregression=> select 'today'::datetime;\n?column?\n----------------------------\nSun Oct 25 00:00:00 1998 EDT\n(1 row)\n\nregression=> select 'tomorrow'::datetime;\n?column?\n----------------------------\nMon Oct 26 00:00:00 1998 EST\n(1 row)\n\nregression=> select 'tomorrow'::datetime - '1 day'::timespan;\n?column?\n----------------------------\nSun Oct 25 01:00:00 1998 EDT\n(1 row)\n\nWhoops. Looks like \"'1 day'::timespan\" is treated as \"24 hours\",\nnot as \"the same local time next day\". Twice a year (if you observe\nDST) there is a difference.\n\nIt's debatable whether this is a bug in the date/time datatypes,\nor just sloppy coding of the regression test. It probably oughta be\ndocumented somewhere, either way. Also, whichever way we decide the\nabove expressions ought to be handled, the docs ought to explain\nhow to get the other functionality.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sun, 25 Oct 1998 19:05:12 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "datetime regression test fails at daylight savings transitions"
},
{
"msg_contents": "> Whoops. Looks like \"'1 day'::timespan\" is treated as \"24 hours\",\n> not as \"the same local time next day\". Twice a year (if you observe\n> DST) there is a difference.\n\nSure. But timespan doesn't know anything about date context. What you\nare really asking about is the behavior with math across two date/time\ndata types.\n\n> It's debatable whether this is a bug in the date/time datatypes,\n> or just sloppy coding of the regression test. It probably oughta be\n> documented somewhere, either way. Also, whichever way we decide the\n> above expressions ought to be handled, the docs ought to explain\n> how to get the other functionality.\n\nHmm. So the debate won't cover whether Tom got it right, just how wrong\nhe got it, eh? :)\n\nThe docs actually mention the distinction between \"qualitative time\"\n(years and months) and \"quantitative time\" (days, hours, minutes,\nseconds) when discussing the \"timespan\" data type.\n\nYou will find that months and years do behave the way you would expect,\nsince they are stored as \"qualitative\" times. So adding a month to a\ntime at the end of the previous month will try to line things up:\n\ntgl=> select '1998-01-31'::datetime + '1 month'::timespan;\n----------------------------\nSat Feb 28 00:00:00 1998 GMT\n\n\nThe real lesson to me is to never put out a \"last call\" to ask people to\nrun the regression tests on the day of a daylight savings time\ntransition...\n\n - Tom\n",
"msg_date": "Mon, 26 Oct 1998 05:34:09 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] datetime regression test fails at daylight savings\n\ttransitions"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n>> Whoops. Looks like \"'1 day'::timespan\" is treated as \"24 hours\",\n>> not as \"the same local time next day\". Twice a year (if you observe\n>> DST) there is a difference.\n\n> Sure. But timespan doesn't know anything about date context. What you\n> are really asking about is the behavior with math across two date/time\n> data types.\n\nRight. The routines that add or subtract a timespan to/from a datetime\nwould have to do this if we want it done.\n\n> Hmm. So the debate won't cover whether Tom got it right, just how wrong\n> he got it, eh? :)\n\nDidn't mean to ruffle your feathers ... the reason that I'm sensitive\nto this issue is that I've blown it several times myself ;-). In the\napplications that I'm working on (financial trading), it's critical that\nthe right things happen at the right times each day ... and the right\ntime is generally defined in terms of local time. Daylight savings time\nbugs have caused us more headaches than I expect to have for the Y2K\nrollover.\n\n> The docs actually mention the distinction between \"qualitative time\"\n> (years and months) and \"quantitative time\" (days, hours, minutes,\n> seconds) when discussing the \"timespan\" data type.\n> You will find that months and years do behave the way you would expect,\n> since they are stored as \"qualitative\" times. So adding a month to a\n> time at the end of the previous month will try to line things up:\n> tgl=> select '1998-01-31'::datetime + '1 month'::timespan;\n> ----------------------------\n> Sat Feb 28 00:00:00 1998 GMT\n\nHmm. This offers a potential solution, then. I propose that \"day\"\nought to be considered a qualitative time interval, and that\n\t'now'::datetime + '1 day'::timespan\nneed not yield the same thing as\n\t'now'::datetime + '24 hours'::timespan\nInstead the first form ought to yield the same local time on the target\nday. (I assume adding months or years does preserve the local time\nof day already?)\n\nChanging things in that way might be infeasible because of backwards\ncompatibility constraints, but I think this is what the natural\ninterpretation would be. (Clearly it's what the writer of the datetime\nregression test was expecting...)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 26 Oct 1998 12:08:00 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] datetime regression test fails at daylight savings\n\ttransitions"
},
{
"msg_contents": "> Hmm. This offers a potential solution, then. I propose that \"day\"\n> ought to be considered a qualitative time interval, and that\n> 'now'::datetime + '1 day'::timespan\n> need not yield the same thing as\n> 'now'::datetime + '24 hours'::timespan\n> Changing things in that way might be infeasible because of backwards\n> compatibility constraints, but I think this is what the natural\n> interpretation would be. (Clearly it's what the writer of the \n> datetime regression test was expecting...)\n\nWell, no I wasn't expecting that really :)\n\nI just wanted to be sure to test 'yesterday' and 'tomorrow' behavior,\nand didn't want to omit those tests just because they might fail for ~1%\nof the year.\n\nMaking 'day' a qualitative time is probably possible, just chewing up\nanother 4 bytes of storage (for 16 bytes rather than 12). But we'll have\nto think it through to make sure there aren't other side effects or\nother no-so-expected behavior under other conditions.\n\n - Tom\n",
"msg_date": "Mon, 26 Oct 1998 17:42:35 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] datetime regression test fails at daylight savings\n\ttransitions"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> Making 'day' a qualitative time is probably possible, just chewing up\n> another 4 bytes of storage (for 16 bytes rather than 12). But we'll have\n> to think it through to make sure there aren't other side effects or\n> other no-so-expected behavior under other conditions.\n\nWell, actually, this is a hard problem, and I don't think it's really\npossible to get it right with a pure data-type-based approach (unless\nyou are willing to have many more and more-complex types than Postgres\nnow does).\n\nTo take one example, if we were to distinguish '24 hours' (absolute) and\n'1 day' (symbolic) in timespans, which flavor should subtracting two\ndatetime values produce? Right now we get stuff like\n\nplay=> select '1998-10-26'::datetime - '1998-10-24'::datetime;\n?column?\n---------------\n@ 2 days 1 hour\n(1 row)\n\nwhich is correct (if \"day\" == \"24 hours\") but probably unexpected.\n\nIn fact it's worse than that, because the behavior depends on\ntimezone setting:\n\nplay=> set timezone = 'GMT';\nSET VARIABLE\nplay=> select '1998-10-26'::datetime - '1998-10-24'::datetime;\n?column?\n--------\n@ 2 days\n(1 row)\n\nHere the timezone is (I presume) being applied during conversion of the\nstring constants into datetime values, while the subtract operator isn't\npaying attention to it. But the operator's behavior would also become\ntimezone-sensitive if it had to produce a \"symbolic\" timespan. I\nsuspect we don't really want operator behavior varying depending on the\ntimezone setting --- it's bad enough that the I/O behavior varies.\n\nI think it's easy to come up with examples of applications that would\nwant either behavior, so you can't expect a single subtract operator\nto solve everyone's problem. Probably best to leave the behavior of\nthe data types alone --- a person who wants the kind of behavior I'm\nthinking of will need to do his own programming. (In practice, you\ncan't do this sort of arithmetic with only timestamps as input anyway;\nyou need context information about timezone, holidays, etc, and you\nprobably want to pass the context info explicitly rather than via a\ncentral global variable. Otherwise dealing with more than one timezone\nis a nightmare.)\n\nSo I withdraw the suggestion that we need to rethink the behavior of\nthe date/time datatypes. But I do suggest that we document somewhere\nthat the datetime regression test will fail near daylight savings\ntransitions... perhaps the README in the test/regress directory is the\nright place. Or could we put a comment into the SQL test code, so that\nit would be visible right in the regression diffs when the problem\noccurs?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 26 Oct 1998 16:44:22 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] datetime regression test fails at daylight savings\n\ttransitions"
}
] |
[
{
"msg_contents": "\nA few things:\n\nSome of the index links on the website are broken. They simply show a\ndirectory listing (Third Party is an example).\n\nAlso, I'm getting emails saying my website is down. I checked the web\nsite, and found that my old domain details are back.\n\nCan someone change them please. The email ones are not that important as\nthe old ones still work, but the web site has been gone since May!\n\nOld\t\t\t\tNew\n\[email protected]\[email protected]\[email protected]\n\nhttp://www.demon.co.uk/finder\thttp://www.retep.org.uk\n\nThanks, peter\n\n-- \n Peter T Mount [email protected]\n Main Homepage: http://www.retep.org.uk\nPostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n Java PDF Generator: http://www.retep.org.uk/pdf\n\n",
"msg_date": "Mon, 26 Oct 1998 06:45:55 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": true,
"msg_subject": "Web site is broken"
}
] |
[
{
"msg_contents": "Hi,\n\nI think we need to add this to INSTALL file after setting\nenvironmnet variable TZ.\n\n'If you compile Postgres with --enable-locale option \n set locale environment to C (or unset all LC_* variables)\n before starting postmaster.'\n\nSeveral regression tests could failed if user's locale collation\nscheme is different from that of standard C locale.\n\nFor example select_implicit and select_having failed on my\nLinux machine when I start postmaster and use koi8-r locale\nand passed fine if I unset all locale variables.\nI don't know why collation scheme of my locale affects to english \ncharacters ? but this could be a source of confusion for people.\n\nresults/select_implicit.out:\nc |count\n--------+-----\nAAAA | 2\nBBBB | 2\nbbbb | 1 ?????\nCCCC | 2\ncccc | 2\nXXXX | 1\n\nexpected/select_implicit.out:\nc |count\n--------+-----\nAAAA | 2\nBBBB | 2\nCCCC | 2\nXXXX | 1\nbbbb | 1\ncccc | 2\n\nI think my locale is broken in this respect. \n\n\n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n",
"msg_date": "Mon, 26 Oct 1998 11:20:39 +0300 (MSK)",
"msg_from": "Oleg Bartunov <[email protected]>",
"msg_from_op": true,
"msg_subject": "a little addition to INSTALL"
}
] |
[
{
"msg_contents": "Hi,\n\nPostgreSQL 6.3.2 was built for Windows\nwith MS VC++ 5.0 and is downloadable from\n\nhttp://www.wipro.com/uwin/ported.html\n\nIt works under UWIN shell developed by \nD. Korn with AT&T research.\nPlease let me know if there is any interest\nin this port. We can probably add a new configure's\ntemplate/target.\n\nThank you\nDmitry Potapov.\n\n\n______________________________________________________\nGet Your Private, Free Email at http://www.hotmail.com\n\n\n",
"msg_date": "Mon, 26 Oct 1998 11:38:52 PST",
"msg_from": "\"Dmitry Potapov\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "PostgreSQL 6.3.2 for Windows NT/95"
}
] |
[
{
"msg_contents": "\nHi all,\n\nAfter a postmaster crash during the vacuum operation, I was unable\nto use my database. I always get the following messages whenever I\ntry to use my database:\n\n LockRelease: locktable lookup failed, no lock\n LockRelease: find xid, table corrupted\n\nIs there a way to get rid of the above messages. Or shoul I forgot \nmy old data and start over from the scrach?\n\nIs there a way to increase the maximum number of locks?\n\nBy the way, I searched the mailin list archives and found that same\nproblem was previously posted for several times, but I could not find any\nsolutions to the problem through the archives.\n\nPlatform info:\n\nOS: IRIX 6.5\nHW: SGI Origin 2000 (4 CPU)\nCompiler used: MIPSPro C\nVersion: PostgreSQL-6-3-2\n\nThanks in advance\n\nHakan Kalyoncu\[email protected]\n\n\n\n",
"msg_date": "Mon, 26 Oct 1998 15:01:46 +0200 (EET)",
"msg_from": "Hakan Kalyoncu <[email protected]>",
"msg_from_op": true,
"msg_subject": "corrupted lock table (fwd)"
}
] |
[
{
"msg_contents": "> >_ FreeBSD 2.2.x-3.x\n> >_ NetBSD 1.3\n> >_ HPUX 9.x\n> >_ HPUX 10.20\n\nHi Stan. I have a recent report of success on NetBSD-1.3.2/x86; don't\nknow which CPU or OS rev you are running and I still need reports for\nNS32532, Sparc, and VAX variants. Also, the HPUX ones have one report\neach (along with some patches to get the hp compiler to work with it;\nthese are now in the main tree).\n\nI haven't gotten a report yet for FreeBSD (I'm guessing that I may see\none from Marc/scrappy), and I'd like to see a second report for the\nother three, but they aren't absolutely essential.\n\nSo, I still need FreeBSD and, depending on your NetBSD rev and\nenvironment, that one too.\n\nTIA\n\n - Tom\n",
"msg_date": "Mon, 26 Oct 1998 14:39:49 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call?"
},
{
"msg_contents": "On Mon, 26 Oct 1998, Thomas G. Lockhart wrote:\n\n> > >_ FreeBSD 2.2.x-3.x\n> > >_ NetBSD 1.3\n> > >_ HPUX 9.x\n> > >_ HPUX 10.20\n> \n> Hi Stan. I have a recent report of success on NetBSD-1.3.2/x86; don't\n> know which CPU or OS rev you are running and I still need reports for\n> NS32532, Sparc, and VAX variants. Also, the HPUX ones have one report\n> each (along with some patches to get the hp compiler to work with it;\n> these are now in the main tree).\n> \n> I haven't gotten a report yet for FreeBSD (I'm guessing that I may see\n> one from Marc/scrappy), and I'd like to see a second report for the\n> other three, but they aren't absolutely essential.\n> \n> So, I still need FreeBSD and, depending on your NetBSD rev and\n> environment, that one too.\n\t\n\tBuilding and doing regression tests as i type...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Mon, 26 Oct 1998 23:06:14 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Last call?"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n\n> Hi Stan. I have a recent report of success on NetBSD-1.3.2/x86; don't\n> know which CPU or OS rev you are running and I still need reports for\n> NS32532, Sparc, and VAX variants.\n\nI'll try to find time to regression test NetBSD/sparc tonight. My VAX\nis having problems right now, so that won't be possible, I'm afraid.\nNo real reason to think it should be broken, but I guess it'll just\nhave to be listed as unsupported, but probably OK.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n",
"msg_date": "27 Oct 1998 09:39:42 +0100",
"msg_from": "Tom Ivar Helbekkmo <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Last call?"
},
{
"msg_contents": "> I'll try to find time to regression test NetBSD/sparc tonight. My VAX\n> is having problems right now, so that won't be possible, I'm afraid.\n> No real reason to think it should be broken, but I guess it'll just\n> have to be listed as unsupported, but probably OK.\n\nI'm planning on leaving the supported machine info the same, so it will\nshow that it was tested under v6.3.x. I'll also have a comment that\nmachines which worked on v6.3 should work on v6.4. But of course it\nwould be best to get new info when possible.\n\n - Tom^H^H^HThomas\n",
"msg_date": "Tue, 27 Oct 1998 15:26:51 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call?"
}
] |
[
{
"msg_contents": "> > _ DGUX 5.4R4.11 m88k\n> Dude, I just lost my DGUX box, so I will no longer be able to verify\n> DGUX ports. Sorry.\n\nOK, thanks for letting me know. You still running Postgres? Especially\non other exotic platforms? Always fishing for new ones... :)\n\nAnyone else out there running DGUX who wants to continue verifying this\nport? I'm pretty sure we are OK for this release, but we don't want to\nget too stale.\n\n - Tom\n",
"msg_date": "Mon, 26 Oct 1998 14:50:59 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Last call?"
}
] |
[
{
"msg_contents": "> > x Windows NT\n\n> How should I list WindowsNT? I have \"mostly working, check \n> web site for\n> patches\" at the moment, but I don't know if we have more info \n> available\n> now.\nYes, it is really \"mostly working ...\" (no dynamic loaded code yet, some\nother unsuccessful regression tests) and there is a website\nhttp://www.askesis.nl/AskesisPostgresIndex.html dedicated to this port. Here\nare all the patches, some porting infos, etc.\n\n\t\t\t\tDan Horak\n",
"msg_date": "Mon, 26 Oct 1998 16:35:39 +0100",
"msg_from": "Horak Daniel <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Last call?"
},
{
"msg_contents": "> > How should I list WindowsNT? I have \"mostly working, check \n> > web site for\n> > patches\" at the moment, but I don't know if we have more info \n> > available\n> > now.\n> Yes, it is really \"mostly working ...\" (no dynamic loaded code yet, some\n> other unsuccessful regression tests) and there is a website\n> http://www.askesis.nl/AskesisPostgresIndex.html dedicated to this port. Here\n> are all the patches, some porting infos, etc.\n\nI have added this to the FAQ:\n\n\tThe database server is now working on Windows NT using the Cygnus\n\tUnix/NT porting library. The only feature missing is dynamic loading of\n\tuser-defined functions/types. See\n\thttp://www.askesis.nl/AskesisPostgresIndex.html for more information. \n\nI have removed the pg_version table, so your pg_ver fix is no longer\nneeded. This should shrink down the patch quite a bit. If it is small\nenough, we can integrate it into the next minor release. Of course, you\nshould continue maintaining the page as you improve the port.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 26 Oct 1998 12:00:18 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Last call?"
}
] |
[
{
"msg_contents": "\nAny objections to my putting out a BETA3 late tomorrow afternoon? \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Mon, 26 Oct 1998 23:10:48 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "BETA3..."
},
{
"msg_contents": "On Mon, 26 Oct 1998, The Hermit Hacker wrote:\n\n> \n> Any objections to my putting out a BETA3 late tomorrow afternoon? \n\nI've just checked cvs, and the files that I reported about 2 weeks ago of\nbeing in the wrong directories are still there.\n\nThis is causing some confusion with a few people when they are trying to\nsee how the driver works.\n\nThe following files need to be deleted to prevent confusion\n\nThe correct one is in the parent directory:\nU src/interfaces/jdbc/postgresql/ChangeLog\n\nThis was moved into the utils sub directory prior to 6.3, and has been\nmodified considerably since then:\n\nU src/interfaces/jdbc/postgresql/PG_Object.java\n\nThese were moved into the geometric sub directory prior to 6.3, and have\nbeen modified considerably since then:\n\nU src/interfaces/jdbc/postgresql/PGbox.java\nU src/interfaces/jdbc/postgresql/PGcircle.java\nU src/interfaces/jdbc/postgresql/PGlseg.java\nU src/interfaces/jdbc/postgresql/PGpath.java\nU src/interfaces/jdbc/postgresql/PGpoint.java\nU src/interfaces/jdbc/postgresql/PGpolygon.java\nU src/interfaces/jdbc/postgresql/PGtokenizer.java\n\nI'm not sure if you are keeping these for historical reasons, but I'm\ngetting some email asking why there are two versions of some of these\nclasses.\n\nPeter\n\n-- \n Peter T Mount [email protected]\n Main Homepage: http://www.retep.org.uk\nPostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n Java PDF Generator: http://www.retep.org.uk/pdf\n\n",
"msg_date": "Tue, 27 Oct 1998 06:39:38 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] BETA3..."
},
{
"msg_contents": "\nI've just checked what I presume is the 6.4 final cvs, and the errant\nfiles are still there... Peter\n\nOn Tue, 27 Oct 1998, Peter T Mount wrote:\n\n> On Mon, 26 Oct 1998, The Hermit Hacker wrote:\n> \n> > \n> > Any objections to my putting out a BETA3 late tomorrow afternoon? \n> \n> I've just checked cvs, and the files that I reported about 2 weeks ago of\n> being in the wrong directories are still there.\n> \n> This is causing some confusion with a few people when they are trying to\n> see how the driver works.\n> \n> The following files need to be deleted to prevent confusion\n> \n> The correct one is in the parent directory:\n> U src/interfaces/jdbc/postgresql/ChangeLog\n> \n> This was moved into the utils sub directory prior to 6.3, and has been\n> modified considerably since then:\n> \n> U src/interfaces/jdbc/postgresql/PG_Object.java\n> \n> These were moved into the geometric sub directory prior to 6.3, and have\n> been modified considerably since then:\n> \n> U src/interfaces/jdbc/postgresql/PGbox.java\n> U src/interfaces/jdbc/postgresql/PGcircle.java\n> U src/interfaces/jdbc/postgresql/PGlseg.java\n> U src/interfaces/jdbc/postgresql/PGpath.java\n> U src/interfaces/jdbc/postgresql/PGpoint.java\n> U src/interfaces/jdbc/postgresql/PGpolygon.java\n> U src/interfaces/jdbc/postgresql/PGtokenizer.java\n> \n> I'm not sure if you are keeping these for historical reasons, but I'm\n> getting some email asking why there are two versions of some of these\n> classes.\n> \n> Peter\n> \n> -- \n> Peter T Mount [email protected]\n> Main Homepage: http://www.retep.org.uk\n> PostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n> Java PDF Generator: http://www.retep.org.uk/pdf\n> \n> \n\n-- \n Peter T Mount [email protected]\n Main Homepage: http://www.retep.org.uk\nPostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n Java PDF Generator: http://www.retep.org.uk/pdf\n\n",
"msg_date": "Sun, 1 Nov 1998 11:30:17 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] BETA3..."
},
{
"msg_contents": "\ndone...\n\nOn Sun, 1 Nov 1998, Peter T Mount wrote:\n\n> \n> I've just checked what I presume is the 6.4 final cvs, and the errant\n> files are still there... Peter\n> \n> On Tue, 27 Oct 1998, Peter T Mount wrote:\n> \n> > On Mon, 26 Oct 1998, The Hermit Hacker wrote:\n> > \n> > > \n> > > Any objections to my putting out a BETA3 late tomorrow afternoon? \n> > \n> > I've just checked cvs, and the files that I reported about 2 weeks ago of\n> > being in the wrong directories are still there.\n> > \n> > This is causing some confusion with a few people when they are trying to\n> > see how the driver works.\n> > \n> > The following files need to be deleted to prevent confusion\n> > \n> > The correct one is in the parent directory:\n> > U src/interfaces/jdbc/postgresql/ChangeLog\n> > \n> > This was moved into the utils sub directory prior to 6.3, and has been\n> > modified considerably since then:\n> > \n> > U src/interfaces/jdbc/postgresql/PG_Object.java\n> > \n> > These were moved into the geometric sub directory prior to 6.3, and have\n> > been modified considerably since then:\n> > \n> > U src/interfaces/jdbc/postgresql/PGbox.java\n> > U src/interfaces/jdbc/postgresql/PGcircle.java\n> > U src/interfaces/jdbc/postgresql/PGlseg.java\n> > U src/interfaces/jdbc/postgresql/PGpath.java\n> > U src/interfaces/jdbc/postgresql/PGpoint.java\n> > U src/interfaces/jdbc/postgresql/PGpolygon.java\n> > U src/interfaces/jdbc/postgresql/PGtokenizer.java\n> > \n> > I'm not sure if you are keeping these for historical reasons, but I'm\n> > getting some email asking why there are two versions of some of these\n> > classes.\n> > \n> > Peter\n> > \n> > -- \n> > Peter T Mount [email protected]\n> > Main Homepage: http://www.retep.org.uk\n> > PostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n> > Java PDF Generator: http://www.retep.org.uk/pdf\n> > \n> > \n> \n> -- \n> Peter T Mount [email protected]\n> Main Homepage: http://www.retep.org.uk\n> PostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n> Java PDF Generator: http://www.retep.org.uk/pdf\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sun, 1 Nov 1998 12:54:09 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] BETA3..."
},
{
"msg_contents": "On Sun, 1 Nov 1998, The Hermit Hacker wrote:\n\n> done...\n\nThanks, it looks a lot neater now.\n\n> \n> On Sun, 1 Nov 1998, Peter T Mount wrote:\n> \n> > \n> > I've just checked what I presume is the 6.4 final cvs, and the errant\n> > files are still there... Peter\n> > \n> > On Tue, 27 Oct 1998, Peter T Mount wrote:\n> > \n> > > On Mon, 26 Oct 1998, The Hermit Hacker wrote:\n> > > \n> > > > \n> > > > Any objections to my putting out a BETA3 late tomorrow afternoon? \n> > > \n> > > I've just checked cvs, and the files that I reported about 2 weeks ago of\n> > > being in the wrong directories are still there.\n> > > \n> > > This is causing some confusion with a few people when they are trying to\n> > > see how the driver works.\n> > > \n> > > The following files need to be deleted to prevent confusion\n> > > \n> > > The correct one is in the parent directory:\n> > > U src/interfaces/jdbc/postgresql/ChangeLog\n> > > \n> > > This was moved into the utils sub directory prior to 6.3, and has been\n> > > modified considerably since then:\n> > > \n> > > U src/interfaces/jdbc/postgresql/PG_Object.java\n> > > \n> > > These were moved into the geometric sub directory prior to 6.3, and have\n> > > been modified considerably since then:\n> > > \n> > > U src/interfaces/jdbc/postgresql/PGbox.java\n> > > U src/interfaces/jdbc/postgresql/PGcircle.java\n> > > U src/interfaces/jdbc/postgresql/PGlseg.java\n> > > U src/interfaces/jdbc/postgresql/PGpath.java\n> > > U src/interfaces/jdbc/postgresql/PGpoint.java\n> > > U src/interfaces/jdbc/postgresql/PGpolygon.java\n> > > U src/interfaces/jdbc/postgresql/PGtokenizer.java\n> > > \n> > > I'm not sure if you are keeping these for historical reasons, but I'm\n> > > getting some email asking why there are two versions of some of these\n> > > classes.\n> > > \n> > > Peter\n> > > \n> > > -- \n> > > Peter T Mount [email protected]\n> > > Main Homepage: http://www.retep.org.uk\n> > > PostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n> > > Java PDF Generator: http://www.retep.org.uk/pdf\n> > > \n> > > \n> > \n> > -- \n> > Peter T Mount [email protected]\n> > Main Homepage: http://www.retep.org.uk\n> > PostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n> > Java PDF Generator: http://www.retep.org.uk/pdf\n> > \n> \n> Marc G. Fournier \n> Systems Administrator @ hub.org \n> primary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n> \n\n-- \n Peter T Mount [email protected]\n Main Homepage: http://www.retep.org.uk\nPostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n Java PDF Generator: http://www.retep.org.uk/pdf\n\n",
"msg_date": "Sun, 1 Nov 1998 18:51:25 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] BETA3..."
}
] |
[
{
"msg_contents": "\nchmod 755 blib/arch/auto/Pg/Pg.so\ncp Pg.bs blib/arch/auto/Pg/Pg.bs\nchmod 644 blib/arch/auto/Pg/Pg.bs\nManifying blib/man3/Pg.3\nmkdir /usr/local/lib/perl5/site_perl: No such file or directory at \\\n\t/usr/libdata/perl/5.00502/ExtUtils/Install.pm line 57\ngmake[3]: *** [pure_site_install] Error 2\n\nneat thing is that I swore that I took out perl5 altogether if the install\nwasn't being performed by root...and looking at the logs, it appears I\ndid, and it got undone:\n\n-----------------------\nrevision 1.214\ndate: 1998/10/13 17:26:40; author: scrappy; state: Exp; lines: +9 -9\n\nchange configure so that if postgresql isn't being installed as root,\ndo not configure in the perl5 interface.\n\nthe perl5 interface needs to be installed under /usr/local/lib/perl5/*,\nwhich is generally owned by root. This allows a non-root build/install\nwith the only root requirement being the make/install of hte perl5\nstuff...\n-------------------------\n\nSo, who took the root check out, and when, and why? \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 02:18:26 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "make install fails in perl5 ..."
},
{
"msg_contents": " neat thing is that I swore that I took out perl5 altogether if the install\n wasn't being performed by root...and looking at the logs, it appears I\n did, and it got undone:\n\n So, who took the root check out, and when, and why? \n\nProbably me, for a couple of reasons.\n\n- It seemed like the lack of a --with-perl option handles this. I can\n see why we might want to verify the --with-perl option and override\n it, but what do we do if the user has perl installed in a way that\n the non-root postgresql installer can write too? Won't this prevent\n a perfectly good installation? For that reason, I prefer requiring\n --with-perl and accepting it. After all, we assume all sorts of\n configure options are ok and don't override them; why is perl\n special in this regard?\n\n- It was also the case that the test involved ${WHOAMI} (I think) but\n no such variable was defined in configure. As a result, the test\n didn't work anyway and always failed. I pointed this out when I\n sent a patch (along with some other changes) and suggested two ways\n to go (delete the test or fix ${WHOAMI}) but I guess no one picked\n up on it and the test got deleted.\n\nI have no problem with putting it back as long as we can address the\ncondition in which perl is writable by nonroot users, but it needs to\ndefine the variables before use. Overall, though I feel that the\nconfigure options are there for the user to provide correct\ninformation, and that this test is not appropriate.\n\nSorry for the confusion.\n\nCheers,\nBrook\n\n",
"msg_date": "Tue, 27 Oct 1998 08:46:39 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "The Hermit Hacker <[email protected]> writes:\n> So, who took the root check out, and when, and why? \n\nWell, I didn't do it, but I agree with whoever took it out. You cannot\ntest at *configure* time to see whether the invoker of configure is\nroot. It is not reasonable to expect people to do the configure and\nbuild as root, even if they su to root for the install step (which\nis surely the WRONG thing for a Postgres install anyway!)\n\nI wouldn't care for a test on rootness even if it were done at the right\ntime, ie, install step. That's an entirely unwarranted assumption about\nhow people set up their file ownership. (For example, on my machine the\nPerl5 tree belongs to user \"gnu\", not root.) A more useful test would\nbe whether you have write permission on the Perl5 install tree top-level\ndirectory, but that seems to require knowing where the Perl5 install\ntree *is*, a fact that's buried inside the Perl-generated makefile.\n\nMaybe we could modify src/interfaces/Makefile along this line:\n\ninstall-perl5:\n\t$(MAKE) -C perl5 clean\n\tcd perl5 && POSTGRES_HOME=\"$(POSTGRESDIR)\" perl Makefile.PL\n\tif test -w `PERL-LIB` then\n\t\t$(MAKE) -C perl5 install\n\t\trm -f perl5/Makefile\n\telse\n\t\techo \"Postponing Perl install, don't have permissions.\"\n\t\techo \"To install, cd into interfaces/perl5, su to Perl\"\n\t\techo \"file owner, and do 'make install'.\"\n\tfi\n\nwhere PERL-LIB is some sed hackery to extract the PERL_LIB\ndefinition from the just-built perl5/Makefile.\n\nIf you like this I will see what I can do with the idea.\n\nWe'll also need to discuss this whole business in the INSTALL\ndirections.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 11:27:24 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "On Tue, 27 Oct 1998, Tom Lane wrote:\n\n> The Hermit Hacker <[email protected]> writes:\n> > So, who took the root check out, and when, and why? \n> \n> Well, I didn't do it, but I agree with whoever took it out. You cannot\n> test at *configure* time to see whether the invoker of configure is\n> root. It is not reasonable to expect people to do the configure and\n> build as root, even if they su to root for the install step (which\n> is surely the WRONG thing for a Postgres install anyway!)\n\nHmmm. That's how I do it, although it *is* a PITA to chown everything\nto do the regression tests. It just seems odd to have one package\ninstall this way when all the other packages I install are done from\nroot.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n\n",
"msg_date": "Tue, 27 Oct 1998 11:47:55 -0500 (EST)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "> Maybe we could modify src/interfaces/Makefile along this line:\n<snip> \n> We'll also need to discuss this whole business in the INSTALL\n> directions.\n\nThis would be OK, but imho it would be acceptable to omit the perl\ninstallation stuff from \"make install\" altogether. Then do the su to\nroot and install explicitly.\n\nOr, have an \"install-root\" or somesuch in top level makefile to have a\ncatchall place for it.\n\nSo,\n\n make install\n su\n make install-root\n\nor something like that.\n\nPer my recent e-mail, we've also got tcl troubles in the installation...\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 17:07:01 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "> Hmmm. That's how I do it, although it *is* a PITA to chown everything\n> to do the regression tests. It just seems odd to have one package\n> install this way when all the other packages I install are done from\n> root.\n\nThere is a security exposure to running the server as root. Other\npackages can make their own security arrangements; this is ours.\n\nMy commercial Ingres package didn't run as root either.\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 17:23:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "On Tue, 27 Oct 1998, Thomas G. Lockhart wrote:\n\n> > Hmmm. That's how I do it, although it *is* a PITA to chown everything\n> > to do the regression tests. It just seems odd to have one package\n> > install this way when all the other packages I install are done from\n> > root.\n> \n> There is a security exposure to running the server as root. Other\n> packages can make their own security arrangements; this is ours.\n\nI understand that, and I *don't* run it as root, I only install it\nas root. After it's installed I chown it to postgres which is what\nmany other packages do. I also do everything from /usr/local/src/\ninstead of /usr/src. IMO, /usr/src is no place for users to be\nputting or building files. It's also alot easier to back up the local\ntree and be done with it. But these are just my preferences. If\nsomeone else wants to build it in / that's their business.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n\n",
"msg_date": "Tue, 27 Oct 1998 12:31:51 -0500 (EST)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "> I understand that, and I *don't* run it as root, I only install it\n> as root. After it's installed I chown it to postgres which is what\n> many other packages do. I also do everything from /usr/local/src/\n> instead of /usr/src. IMO, /usr/src is no place for users to be\n> putting or building files. It's also alot easier to back up the local\n> tree and be done with it. But these are just my preferences. If\n> someone else wants to build it in / that's their business.\n\nSorry about that. I agree with you completely (I may even agree with you\nmore than you do! :)\n\nI do a _lot_ of postgres building, since developing the server is one of\nmy hobbies. istm it is much cleaner to do everything out of\n/opt/postgres, owned by postgres (as opposed to scattering things below\n/, which as you point out can be a mess). I put the builds into\ndirectories like v6.3.2 or 981014.d, then point to one with a softlink\n/opt/postgres/current. Then nothing needs to happen as root for months\non end, and I can switch between versions in a minute or two if\nnecessary.\n\nI'm not sure what the current thinking is on the /opt concept wrt the\nnewer file system conventions being negotiated between Unices...\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 18:21:54 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": " Well, I didn't do it, but I agree with whoever took it out. You cannot\n test at *configure* time to see whether the invoker of configure is\n root. It is not reasonable to expect people to do the configure and\n build as root, even if they su to root for the install step (which\n is surely the WRONG thing for a Postgres install anyway!)\n\nInstalling as root is fine; running as root is not.\n\n I wouldn't care for a test on rootness even if it were done at the right\n time, ie, install step. That's an entirely unwarranted assumption about\n how people set up their file ownership. (For example, on my machine the\n Perl5 tree belongs to user \"gnu\", not root.) A more useful test would\n be whether you have write permission on the Perl5 install tree top-level\n directory, but that seems to require knowing where the Perl5 install\n tree *is*, a fact that's buried inside the Perl-generated makefile.\n\nWe don't really need to test this. Use --with-perl if appropriate.\nIt will automatically find the right stuff via perl.\n\n Maybe we could modify src/interfaces/Makefile along this line:\n\nNo. Why can't we rely on people using --with-perl correctly to\nspecify if they do or do not want perl. The whole point of the recent\nrun-around with the perl Makefile stuff was so that postgres could be\ninstalled without perl and later someone could do the normal perl\ninstall in that directory. You get this functionality by running\nconfigure without the --with-perl option and all is well.\n\nWhat am I missing?\n\nCheers,\nBrook\n",
"msg_date": "Tue, 27 Oct 1998 11:23:33 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "Vince Vielhaber <[email protected]> writes:\n> I understand that, and I *don't* run it as root, I only install it\n> as root. After it's installed I chown it to postgres which is what\n> many other packages do.\n\nI dunno, that sure seems like the hard way to me. I made\n/usr/local/pgsql (which I had to do as root, since /usr/local isn't\nwidely writable) and chowned it to postgres. Since then, I can do\nthe builds, installs, and everything else as postgres.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 13:29:04 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "Brook Milligan <[email protected]> writes:\n> No. Why can't we rely on people using --with-perl correctly to\n> specify if they do or do not want perl. The whole point of the recent\n> run-around with the perl Makefile stuff was so that postgres could be\n> installed without perl and later someone could do the normal perl\n> install in that directory. You get this functionality by running\n> configure without the --with-perl option and all is well.\n\n> What am I missing?\n\nWhat you're missing is that that renders configure --with-perl \npractically useless.\n\nIf you say --with-perl, that not only builds the perl module along with\npostgres (cool, especially if we ever arrange for pre-install testing),\nbut also tries to *install* it along with postgres (not so cool).\nThe permissions required to do the install may be different for postgres\nitself and a perl module. They *are* different if you follow the\nrecommended install procedure of making postgres be an unprivileged\nuser.\n\nI'm with Marc on this one, actually; I've been hacking around this\npermissions problem by chmod'ing my perl tree when I was about to run\na postgres install, and that's not cool. Becoming root in order to\ninstall is a band-aid, not a proper solution.\n\nAfter I sent off my previous message I realized that it still wasn't\nquite right; the install-perl Makefile rule should rebuild the perl\nmodule with \"make all\" even if it doesn't have permissions to install\nthe result. That way, if I come along and su to the Perl file owner\nand do make install in interfaces/perl5, I don't need write permission\non the perl module files (which are still postgres's). In a setup like\nmine, where the Perl file owner is not root but another unprivileged\naccount much like postgres, it should be possible to do all these\ninstalls without ever becoming root (except to make the initial\n/usr/local/pgsql directory).\n\nI'll fix this this afternoon, unless I hear a better objection...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 13:43:52 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "Tom Lane wrote:\n> \n> Brook Milligan <[email protected]> writes:\n> > No. Why can't we rely on people using --with-perl correctly to\n> > specify if they do or do not want perl. The whole point of the recent\n> > run-around with the perl Makefile stuff was so that postgres could be\n> > installed without perl and later someone could do the normal perl\n> > install in that directory. You get this functionality by running\n> > configure without the --with-perl option and all is well.\n> \n> > What am I missing?\n> \n> What you're missing is that that renders configure --with-perl\n> practically useless.\n> \n> If you say --with-perl, that not only builds the perl module along with\n> postgres (cool, especially if we ever arrange for pre-install testing),\n> but also tries to *install* it along with postgres (not so cool).\n> The permissions required to do the install may be different for postgres\n> itself and a perl module. They *are* different if you follow the\n> recommended install procedure of making postgres be an unprivileged\n> user.\n> \n> I'm with Marc on this one, actually; I've been hacking around this\n> permissions problem by chmod'ing my perl tree when I was about to run\n> a postgres install, and that's not cool. Becoming root in order to\n> install is a band-aid, not a proper solution.\n> \n> After I sent off my previous message I realized that it still wasn't\n> quite right; the install-perl Makefile rule should rebuild the perl\n> module with \"make all\" even if it doesn't have permissions to install\n> the result. That way, if I come along and su to the Perl file owner\n> and do make install in interfaces/perl5, I don't need write permission\n> on the perl module files (which are still postgres's). In a setup like\n> mine, where the Perl file owner is not root but another unprivileged\n> account much like postgres, it should be possible to do all these\n> installs without ever becoming root (except to make the initial\n> /usr/local/pgsql directory).\n> \n> I'll fix this this afternoon, unless I hear a better objection...\n\nI think you are headed in this direction, but istm that the --with-xxx\nflags on configure should work for any package at any time. So someone\ndoing an installation can look at the \"./configure --help\" output and\nsay \"I think I want those\", and actually have it do the right thing,\nstopping before it does the wrong thing. So, --with-perl should enable\nperl building (by just setting USE_PERL or something?) but stop short of\nrequiring root access to the best installation location during the main\nPostgres install.\n\nI should be able to override the \"--with\" settings from configure with\nmy Makefile.custom, so I can reinstall the entire Postgres package just\nby copying Makefile.custom and rerunning configure without major command\nline parameters.\n\nI suspect that the tcl/tk stuff doesn't quite work this way at the\nmoment, and that the perl stuff needs a bit of adjustment also. I'll\nlook at the ODBC stuff to get it in line if it isn't already.\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 19:19:51 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "\nOn 27-Oct-98 Tom Lane wrote:\n> Vince Vielhaber <[email protected]> writes:\n>> I understand that, and I *don't* run it as root, I only install it\n>> as root. After it's installed I chown it to postgres which is what\n>> many other packages do.\n> \n> I dunno, that sure seems like the hard way to me. I made\n> /usr/local/pgsql (which I had to do as root, since /usr/local isn't\n> widely writable) and chowned it to postgres. Since then, I can do\n> the builds, installs, and everything else as postgres.\n\n$ ./configure\n$ gmake\n$ sudo gmake install\n\nMany packages build/install similarly and the makefile ensures that the\npermissions and ownerships are correct. Since the postgresql makefile\ndoesn't do the permissions/ownerships I have to. This last part must\nbe what you see as \"the hard way\". It's what I see as odd (as I stated\nearlier). That's really the only difference. If I don't do the regression\ntests (and I usually don't) I don't have to bother with the permissions \non the source tree.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Tue, 27 Oct 1998 15:12:29 -0500 (EST)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> I think you are headed in this direction, but istm that the --with-xxx\n> flags on configure should work for any package at any time. So someone\n> doing an installation can look at the \"./configure --help\" output and\n> say \"I think I want those\", and actually have it do the right thing,\n> stopping before it does the wrong thing.\n\nRight, my thoughts exactly. A reasonable person seeing the --with-perl\nconfigure option would assume that he wants to use it, if he has Perl.\nAs things stand, if he's following the recommended install procedure\nhe probably can't use it. That's wrong.\n\n> I should be able to override the \"--with\" settings from configure with\n> my Makefile.custom, so I can reinstall the entire Postgres package just\n> by copying Makefile.custom and rerunning configure without major command\n> line parameters.\n\nWell, that should be true as long as you also get the subsidiary\ninformation that configure computes (locations of tclsh and friends, in\nthe Tcl case). It is legitimate for configure to compute those because\nthey are config info --- what if you have more than one version of Tcl\ninstalled? You need to be able to specify which one to use.\n\nWe have two problems here:\n\n1. Makefile.global ought to make clear which other settings need to be\n made if you want to turn on USE_TCL/USE_TK. This is just\n documentation.\n\n2. Setting those other settings in Makefile.custom ought to be\n sufficient. It is not, right now, because configure writes the\n values into non-makefiles that will not notice what you did in\n Makefile.custom. This is fixable but I am leery of trying to fix it\n for 6.4.\n\nI suggest that at this late stage we ought to concentrate on making\nsure that the standard configure/install procedure works smoothly.\nIt's too late to worry about whether people's existing Makefile.custom\nwill work smoothly --- and I'm not sure it's reasonable to expect a\nMakefile.custom to work just because it worked with some prior version,\nanyway.\n\n> I suspect that the tcl/tk stuff doesn't quite work this way at the\n> moment, and that the perl stuff needs a bit of adjustment also. I'll\n> look at the ODBC stuff to get it in line if it isn't already.\n\nI'm not touching ODBC :-).\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 15:24:26 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "> Well, that should be true as long as you also get the subsidiary\n> information that configure computes (locations of tclsh and friends, \n> in the Tcl case). It is legitimate for configure to compute those \n> because they are config info --- what if you have more than one \n> version of Tcl installed? You need to be able to specify which one to \n> use.\n\nSure. But if you have only one version installed, and if things are\nlocated in about the right places, then configure should probably get\nthose things done even if \"--with-tcl\" is not specified. It would at the\nsame time set \"USE_TCL\" to false, which could be overridden later.\n\n> I suggest that at this late stage we ought to concentrate on making\n> sure that the standard configure/install procedure works smoothly.\n> It's too late to worry about whether people's existing Makefile.custom\n> will work smoothly --- and I'm not sure it's reasonable to expect a\n> Makefile.custom to work just because it worked with some prior \n> version, anyway.\n\nI agree, and wasn't advocating that my old Makefile.custom must continue\nto work unchanged. It just helped me notice the other problems sooner by\nhanging my usual, \"should work easy\" installation from a clean tree. And\nagain, it is sensitive to being a pristine installation; after the fact\nit sort of repairs itself.\n\n> > I suspect that the tcl/tk stuff doesn't quite work this way at the\n> > moment, and that the perl stuff needs a bit of adjustment also. I'll\n> > look at the ODBC stuff to get it in line if it isn't already.\n> I'm not touching ODBC :-)\n\n... anymore :) We currently have ODBC breakage in moving to the\nMakefile.shlib. But we've already discussed fixing that. Should I just\ngive it a try, or do you want to see patches?\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 21:00:30 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ..."
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n>> I'm not touching ODBC :-)\n\n> ... anymore :) We currently have ODBC breakage in moving to the\n> Makefile.shlib. But we've already discussed fixing that. Should I just\n> give it a try, or do you want to see patches?\n\nHack away. I'll try to take a look when you're done.\n\nBTW, I believe I have fixed the tcl problem, if you care to try \nthat again.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 17:08:21 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "OK, I've fixed the perl5 permissions problem and verified that it\nactually works right :-). I didn't check in any changes to the\ninstall.sgml file, but suggest you add something like this\nright after the \"make install\" step:\n\n\n<Step Performance=\"optional\">\n<Para>\nIf you used the <option>--with-perl</option> option to configure, check\nthe install log to see whether the Perl module was actually installed.\nIf you've followed our advice to make the Postgres files be owned by\nan unprivileged userid, then the Perl module won't have been installed,\nfor lack of write privileges on the Perl library directories. You can\ncomplete its installation, either now or later, by su'ing to the user that\ndoes own the Perl library and doing\n<ProgramListing>\n$ cd /usr/src/pgsql/src/interfaces/perl5\n$ gmake install\n</ProgramListing>\n</Para>\n</Step>\n\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 18:19:12 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "Tom Lane wrote:\n> The Hermit Hacker <[email protected]> writes:\n> > So, who took the root check out, and when, and why? \n> \n> Well, I didn't do it, but I agree with whoever took it out. You cannot\n> test at *configure* time to see whether the invoker of configure is\n> root. It is not reasonable to expect people to do the configure and\n> build as root, even if they su to root for the install step (which\n> is surely the WRONG thing for a Postgres install anyway!)\n\nI don't know. I install as root (so that the perl stuff will install). Of \ncourse I have the following lines in Makefile.custom:\n\n\tINSTLOPTS += -o bin -g bin\n\tINSTL_EXE_OPTS += -o pgsql -g dbadmin\n\tINSTL_LIB_OPTS += -o bin -g bin\n\tINSTL_SHLIB_OPTS += -o bin -g bin\n\nso that the permissions are set the way I want them and the postgres \nexecutables are owned by the postgres user.\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Thu, 29 Oct 1998 00:42:08 -0500",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
},
{
"msg_contents": "Tom Lane wrote:\n> \"Thomas G. Lockhart\" <[email protected]> writes:\n> > I think you are headed in this direction, but istm that the --with-xxx\n> > flags on configure should work for any package at any time. So someone\n> > doing an installation can look at the \"./configure --help\" output and\n> > say \"I think I want those\", and actually have it do the right thing,\n> > stopping before it does the wrong thing.\n> \n> Right, my thoughts exactly. A reasonable person seeing the --with-perl\n> configure option would assume that he wants to use it, if he has Perl.\n> As things stand, if he's following the recommended install procedure\n> he probably can't use it. That's wrong.\n> \n> > I should be able to override the \"--with\" settings from configure with\n> > my Makefile.custom, so I can reinstall the entire Postgres package just\n> > by copying Makefile.custom and rerunning configure without major command\n> > line parameters.\n\nThen why use configure at all? Configure is there to do a specific job. \nConfiguring PostgreSQL to build and install on your machine. Use it to do \nthat job.\n\n> \n> Well, that should be true as long as you also get the subsidiary\n> information that configure computes (locations of tclsh and friends, in\n> the Tcl case). It is legitimate for configure to compute those because\n> they are config info --- what if you have more than one version of Tcl\n> installed? You need to be able to specify which one to use.\n\nThere are options in configure to let you specify the directory that contains \nthe tclConfig.sh and tkConfig.sh. This overides the normal search sequence \nconfigure uses to find the *Config.sh files.\n\n> \n> We have two problems here:\n> \n> 1. Makefile.global ought to make clear which other settings need to be\n> made if you want to turn on USE_TCL/USE_TK. This is just\n> documentation.\n\nThe documentation should make it clear that the Makefile.global file (or any \nother file that configure generates) should not be edited by hand. Configure \nshould be used to change the configuration options!\n\nOf course, this is just my opinion, and like any other philosophical stand so \ngenerate interesting debates :-)\n\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Thu, 29 Oct 1998 01:10:32 -0500",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] make install fails in perl5 ... "
}
] |
[
{
"msg_contents": "I started out with a Makefile.custom which has in it the following:\n\nUSE_TCL= true\nTCL_LIB= -ltcl\nX_LIBS= -L/usr/X11/lib\nTK_LIB= -ltk\n\nDid a \"configure\", without specifying --with-tcl, then did a \"make\ninstall\". The installation process hangs forever at the\n\"mkMakefile.tcldefs.sh\" stage:\n\nmake[2]: Entering directory `/opt/postgres/pgsql/src/bin/pgtclsh'\nMakefile:20: Makefile.tcldefs: No such file or directory\n/bin/sh mkMakefile.tcldefs.sh\nmake[2]: *** Deleting file `Makefile.tcldefs'\nmake[2]: *** [Makefile.tcldefs] Interrupt\nmake[1]: *** [install] Interrupt\nmake: *** [install] Interrupt\n\nDid a ^C to exit. Then, I reran configure with --with-tcl specified this\ntime, and from the bin/pgtclsh directory did a \"make\":\n\ngolem$ make\nMakefile:20: Makefile.tcldefs: No such file or directory\nMakefile:22: Makefile.tkdefs: No such file or directory\n/bin/sh mkMakefile.tkdefs.sh\n/bin/sh mkMakefile.tcldefs.sh\ngcc -I../../include -I../../backend -O2 -m486 -Wall\n-Wmissing-prototypes -I/usr/X11R6/include -I../../interfaces/libpgtcl \n-c pgtclAppInit.c -o pgtclAppInit.o\n\nIt looks like it now made the file which is required by the Makefile\n(!?), so retrying \"make\" seems to run to completion.\n\nI know there was quite a bit of discussion on the hackers list about how\nto do this installation, but it doesn't seem quite right or at lesat\nfinished (though I know that the \"USE_TCL\" in my Makefile.custom might\nhave thrown it off a little bit). Did this result in any written docs on\nhow to install? Is someone prepared to fix up the Makefile to make it\nmore robust (e.g. it can't include a file which hasn't been constructed\nyet, though there are ways around that by conditionally including it and\nthen running make from within that makefile).\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 14:48:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "TCL installation troubles"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> I started out with a Makefile.custom which has in it the following:\n> USE_TCL= true\n> TCL_LIB= -ltcl\n> X_LIBS= -L/usr/X11/lib\n> TK_LIB= -ltk\n> [ and it didn't work ]\n\nThere is a comment in Makefile.global.in that says:\n\t# Please do not edit USE_TCL and USE_TK by hand.\nI dunno who put that in or why, but it seems relevant ;-)\n\nAn offhand look at configure.in shows that it is also deriving values\nfor TCL_CONFIG_SH and TK_CONFIG_SH, which need to be propagated into\nplaces that Makefile.custom can't reach. It looks to me like\nmkMakefile.tcldefs.sh is probably not coping gracefully if an empty\nstring is substituted into it, which is what will happen right now\nif USE_TCL isn't set during configure. By setting USE_TCL in\nMakefile.custom, you made the Makefiles think that Tcl support is\nset up, but that doesn't take care of the subsidiary files.\n\n> Is someone prepared to fix up the Makefile to make it\n> more robust (e.g. it can't include a file which hasn't been constructed\n> yet, though there are ways around that by conditionally including it and\n> then running make from within that makefile).\n\nWell, *that* is not the problem here. GNU Make is smarter than you\nrealize about files that are included into a makefile --- it will\narrange to build the file if missing and then reread the makefile.\nThe problem is an inconsistent set of configuration values caused\nby your hand-setting only some of the needed variables. And you can't\neven fix it by adding the missing variables to Makefile.custom.\n\n\nThis brings up an issue that I had been planning to raise in response\nto Brook and Billy's disagreement a couple days ago about how to handle\n\"libdir\" in the Makefiles (thread \"Configure problem (and fix)\").\nNamely: I think we have gotten much too willing to use configure to\nrewrite subsidiary files all over the distribution, rather than ensuring\nthat the configuration decisions are expressed in a central place.\n\nI believe it is considered good Autoconf style to emit a Makefile (or\nwhatever) that is still hand-editable, so that one can go in and tweak\nthe configure script's decisions after the fact. (In other words,\nBrook had the right idea as to style: libdir should be left unexpanded\nin all references, so that you can actually change it by hand in\nMakefile.global if you are so minded.)\n\nWe have broken that capability entirely --- if you don't set a parameter\nto configure, you have little choice but to go and re-run configure,\nbecause chasing down all of the twenty or so output files that it\ngenerates is not very practical. (Which was Billy's point, or part of\nit; and he's right too.)\n\nThis is not good, and it is not necessary. AFAICS there is *no* good\nreason that configure should be rewriting any subdirectory Makefile ---\nthey all include Makefile.global, so anything they need to know could\nbe stated, once, in Makefile.global. With a little more work we could\nprobably make Makefile.global and config.h be the only files that\nconfigure creates at all. The various specialized shell scripts that\nconfigure is currently editing could be fixed so that they get their\nconfiguration info from the command line, whence it could be supplied\nfrom Makefile.global via the subdirectory Makefile that is invoking the\nscript.\n\nI'm not certain how best to handle the SQL scripts that need to know\nwhere libdir is, but I can think of several possibilities, one being\nthat \"create function\" could have a library search path built into it,\nthus pushing the knowledge of where libdir is into some C code (which\nwould probably be getting it from the PGLIB environment variable).\nOr we could put the value of libdir into a system table somewhere that\nthe scripts can read at runtime.\n\n\nIt's awfully late to be fixing this stuff for 6.4, unless you want to\nslip the release date again. But I suggest revisiting it for 6.4.1,\nand trying to consolidate configuration decisions into as few files\nas possible.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 13:25:12 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Configure problem, redux (was Re: TCL installation troubles)"
},
{
"msg_contents": " I'm not certain how best to handle the SQL scripts that need to know\n where libdir is, but I can think of several possibilities, one being\n that \"create function\" could have a library search path built into it,\n thus pushing the knowledge of where libdir is into some C code (which\n would probably be getting it from the PGLIB environment variable).\n Or we could put the value of libdir into a system table somewhere that\n the scripts can read at runtime.\n\nThe way to handle this is to have rules in the Makefile that do the\nsubstitution. For example, something like the following Makefile\nfragment will do the trick, even if the definitition of $libdir in\nMakefile.global is modified after configure is run.\n\n SRCDIR=../../..\n include ${SRCDIR}/Makefile.global\n mklang.sql: mklang.sql.in ${SRCDIR}/Makefile.global\n\t\t sed < $< > $@ -e 's:@libdir@:${libdir}:'\n\nIf I remember correctly, this is used in contrib/spi and should be\ncommon practice throughout.\n\nYour suggestion of a single central Makefile.global/config.h to\ncontain all the configure decisions is a good one. Together with\nrules such as above in the various appropriate places, the problems of\nvariable expansion and post-configure modifications disappear.\nPerhaps not for 6.4, though.\n\nCheers,\nBrook\n",
"msg_date": "Tue, 27 Oct 1998 11:42:35 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Configure problem, redux (was Re: TCL installation troubles)"
},
{
"msg_contents": "All good points. I had the USE_TCL in my Makefile.custom because for the\nlast few months/years that was the only way to get anything about tcl\ntouched by the installation afaik. I'll take it out.\n\nBut I'm pretty sure that doesn't explain all the breakage. Will continue\ntesting a bit (I *really* need to get back to the docs!), but the first\nproblem I saw was due to a missing file which was not built\nautomatically, and if you don't do a clean install you won't see the\nproblem again. That's why I could work with \"cvs update -Pd\" for weeks\nand not see the breakage introduced, because by that point a\nMakefile.tcldefs already existed.\n\n> It's awfully late to be fixing this stuff for 6.4, unless you want to\n> slip the release date again. But I suggest revisiting it for 6.4.1,\n> and trying to consolidate configuration decisions into as few files\n> as possible.\n\nI would strongly suggest that we fix it now, with small incremental\nchanges to make it work as currently designed. Releasing it broken\ndoesn't do much good.\n\nSo, I'm not sure I understand what the current design is really supposed\nto do, but istm that we could do a conditional include of\nMakefile.tcldefs, and have Makefile.tcldefs be a prerequisite for\n$(PGMS). Like this:\n\n ifneq ($(wildcard Makefile.tclsh), )\n include Makefile.tclsh\n endif\n ...\n all: Makefile.tclsh\n $(MAKE) $(PGMS)\n ...\n\nAny redesign after v6.4 is released should probably wait for v6.5, since\na v6.4.1 release won't get adequate testing. We managed to break the\nLinux port of Postgres for v6.3.1 for similar reasons.\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 19:06:30 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Configure problem, redux (was Re: TCL installation troubles)"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> But I'm pretty sure that doesn't explain all the breakage. Will continue\n> testing a bit (I *really* need to get back to the docs!), but the first\n> problem I saw was due to a missing file which was not built\n> automatically, and if you don't do a clean install you won't see the\n> problem again. That's why I could work with \"cvs update -Pd\" for weeks\n> and not see the breakage introduced, because by that point a\n> Makefile.tcldefs already existed.\n\nAh, I see the problem! The Makefile in .../pgtclsh has\n\nMakefile.tcldefs:\n\t/bin/sh mkMakefile.tcldefs.sh\n\nwhich is OK for building Makefile.tcldefs initially ... but it needs\na dependency so that Makefile.tcldefs will be rebuilt if\nmkMakefile.tcldefs.sh has changed.\n\nI think there is also a syntax error at the inclusion point ---\ntabs before an include directive are a NO-NO.\n\nWill fix, and check the other tcl makefiles for the same problem.\n\n> So, I'm not sure I understand what the current design is really supposed\n> to do, but istm that we could do a conditional include of\n> Makefile.tcldefs, and have Makefile.tcldefs be a prerequisite for\n> $(PGMS). Like this:\n\n> ifneq ($(wildcard Makefile.tclsh), )\n> include Makefile.tclsh\n> endif\n\nLike I said, this is *not* necessary since we are assuming GNU Make.\nMake already knows about (re)building included files as required;\nsee \"How Makefiles Are Remade\" in the GNU Make manual.\nThe problem is we're not giving it correct dependency info.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 15:10:56 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "> \"Thomas G. Lockhart\" <[email protected]> writes:\n> > I started out with a Makefile.custom which has in it the following:\n> > USE_TCL= true\n> > TCL_LIB= -ltcl\n> > X_LIBS= -L/usr/X11/lib\n> > TK_LIB= -ltk\n> > [ and it didn't work ]\n> \n> There is a comment in Makefile.global.in that says:\n> \t# Please do not edit USE_TCL and USE_TK by hand.\n> I dunno who put that in or why, but it seems relevant ;-)\n> \n> An offhand look at configure.in shows that it is also deriving values\n> for TCL_CONFIG_SH and TK_CONFIG_SH, which need to be propagated into\n> places that Makefile.custom can't reach. It looks to me like\n> mkMakefile.tcldefs.sh is probably not coping gracefully if an empty\n> string is substituted into it, which is what will happen right now\n> if USE_TCL isn't set during configure. By setting USE_TCL in\n> Makefile.custom, you made the Makefiles think that Tcl support is\n> set up, but that doesn't take care of the subsidiary files.\n\nOur tcl/tk suff was terrible in earlier releases. The new code from\nBilly is quite nice. It looks at things the way tcl does. It looks for\ntclConfig.sh in all the normal places, and gets the values from there.\n\nWe used to set those values manually, and had all sorts of tcl\nversion-specific tests that were certain to fail on many platforms.\n\n\n> > Is someone prepared to fix up the Makefile to make it\n> > more robust (e.g. it can't include a file which hasn't been constructed\n> > yet, though there are ways around that by conditionally including it and\n> > then running make from within that makefile).\n> \n> Well, *that* is not the problem here. GNU Make is smarter than you\n> realize about files that are included into a makefile --- it will\n> arrange to build the file if missing and then reread the makefile.\n> The problem is an inconsistent set of configuration values caused\n> by your hand-setting only some of the needed variables. And you can't\n> even fix it by adding the missing variables to Makefile.custom.\n> \n> \n> This brings up an issue that I had been planning to raise in response\n> to Brook and Billy's disagreement a couple days ago about how to handle\n> \"libdir\" in the Makefiles (thread \"Configure problem (and fix)\").\n> Namely: I think we have gotten much too willing to use configure to\n> rewrite subsidiary files all over the distribution, rather than ensuring\n> that the configuration decisions are expressed in a central place.\n> \n> I believe it is considered good Autoconf style to emit a Makefile (or\n> whatever) that is still hand-editable, so that one can go in and tweak\n> the configure script's decisions after the fact. (In other words,\n> Brook had the right idea as to style: libdir should be left unexpanded\n> in all references, so that you can actually change it by hand in\n> Makefile.global if you are so minded.)\n> \n> We have broken that capability entirely --- if you don't set a parameter\n> to configure, you have little choice but to go and re-run configure,\n> because chasing down all of the twenty or so output files that it\n> generates is not very practical. (Which was Billy's point, or part of\n> it; and he's right too.)\n> \n> This is not good, and it is not necessary. AFAICS there is *no* good\n> reason that configure should be rewriting any subdirectory Makefile ---\n> they all include Makefile.global, so anything they need to know could\n> be stated, once, in Makefile.global. With a little more work we could\n> probably make Makefile.global and config.h be the only files that\n> configure creates at all. The various specialized shell scripts that\n> configure is currently editing could be fixed so that they get their\n> configuration info from the command line, whence it could be supplied\n> from Makefile.global via the subdirectory Makefile that is invoking the\n> script.\n\nI disagree here. If people want to twiddle, they can after configure. \nPeople would much rather spell params to configure, rather than to edit\nMakefiles.\n\nThe use of configure as a centralized test mechanism is a great leap\nforward for us. The basic problem is that Makefile's just can't do the\ntests configure can, and any Makefile that needs a test needs a\nMakefile.in, and is generated from configure.\n\nLook in Makefile.global. It is huge, and very hard to understand, even\nfor veteran developers. I agree people should be able to twiddle with\nMakefile.custom, but they MUST be required to supply the proper flags to\nconfigure. We can't go down the road of allowing them to avoid the\n'configure' flags, and somehow enable things in Makefile.custom. There\nis no logical way to do that, and unless we want to prevent 'configure'\nfrom doing the things it does so well, we will have to live with that\nlimitation.\n\nI have my configure command and flags in a script and use run it as part\nof the cvs update I do to keep my sources current.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 15:37:14 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "> I'm not certain how best to handle the SQL scripts that need to know\n> where libdir is, but I can think of several possibilities, one being\n> that \"create function\" could have a library search path built into it,\n> thus pushing the knowledge of where libdir is into some C code (which\n> would probably be getting it from the PGLIB environment variable).\n> Or we could put the value of libdir into a system table somewhere that\n> the scripts can read at runtime.\n> \n> The way to handle this is to have rules in the Makefile that do the\n> substitution. For example, something like the following Makefile\n> fragment will do the trick, even if the definitition of $libdir in\n> Makefile.global is modified after configure is run.\n> \n> SRCDIR=../../..\n> include ${SRCDIR}/Makefile.global\n> mklang.sql: mklang.sql.in ${SRCDIR}/Makefile.global\n> \t\t sed < $< > $@ -e 's:@libdir@:${libdir}:'\n\nThe problem here is that you are duplicating the normal configure\nprocessing in every Makefile that needs it. This will get old very\nfast, and hard to maintain. configure does this already, and\nautomatically, in one place. Yes, you must re-run configure, and you do\nloose your changes, but pulling all the stuff into every Makefile seems\nworse.\n\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 15:40:54 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": " The problem here is that you are duplicating the normal configure\n processing in every Makefile that needs it. This will get old very\n fast, and hard to maintain. configure does this already, and\n automatically, in one place. Yes, you must re-run configure, and you do\n loose your changes, but pulling all the stuff into every Makefile seems\n worse.\n\nWe are only doing this for the *.sql files because they don't have\nwithin them some ability to do argument expansion. This is definitely\nnot good for all the Makefiles or source code, both of which can get\nwhat they need in a single place (Makefile.global or config.h). I am\nnot advocating changing that. In fact, I am agreeing with the idea of\nbecoming more centralized for those parts of the code that can handle\nit, i.e., Makefiles and C source.\n\nFor *.sql files there are two choices:\n\n- have configure do the expansion. This requires either special\n variables or expansion of lots of variables, neither of which is a\n good solution because they require unusual coding practices or\n prevent easy maintenance (i.e., change of Makefile.global) later.\n\n- have make do the expansion. Certainly this puts some of configure's\n job into some Makefiles, but it offers the advantage that\n Makefile.global can be changed and those changes will propagate\n where they are needed.\n\nIt seems like we need a well-documented policy decision, perhaps\nbetween the two following options (plus any others people want to\nformulate).\n\n1. we are only speaking (to my knowledge) of the library locations, so\n use ${expanded_libdir} in *.sql files; configure will do the right\n thing with these (i.e., only expand this but leave the \"normal\"\n variables unexpanded so that changes to Makefile.global propagate\n easily), but changes to Makefile.global cannot touch the configured\n choice for ${expanded_libdir}. Changes to that require going\n through all the *.sql files seeking those with the wrong\n expansions (or reconfiguring).\n\n2. add rules to those few Makefiles that require it so that make will\n expand for *.sql based on the configured choices. Changes to\n Makefile.global will do the right thing.\n\nThe latter perhaps duplicates a tiny bit of configure, but is more\nflexible. If the rule is made generic (which it really is) it can\neven be put into Makefile.global as a pattern target and not even\nworried about in any of the Makefiles; if there is a *.sql.in it will\nbe made into *.sql with the correct substitution, end of story, no\nneed to be concerned with duplicating parts of configure all over the\nplace. This follows the recent trend for a centralized set of rules\nfor shared library handling. Why not a central place for the common\n*.sql substitution as well, so that what is in Makefile.global gets\npropagated?\n\nCheers,\nBrook\n",
"msg_date": "Tue, 27 Oct 1998 14:27:25 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "> The problem here is that you are duplicating the normal configure\n> processing in every Makefile that needs it. This will get old very\n> fast, and hard to maintain. configure does this already, and\n> automatically, in one place. Yes, you must re-run configure, and you do\n> loose your changes, but pulling all the stuff into every Makefile seems\n> worse.\n> \n> We are only doing this for the *.sql files because they don't have\n> within them some ability to do argument expansion. This is definitely\n> not good for all the Makefiles or source code, both of which can get\n> what they need in a single place (Makefile.global or config.h). I am\n> not advocating changing that. In fact, I am agreeing with the idea of\n> becoming more centralized for those parts of the code that can handle\n> it, i.e., Makefiles and C source.\n\nOh. :-)\n\nI will go back to looking the other way.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 16:33:39 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n>> Namely: I think we have gotten much too willing to use configure to\n>> rewrite subsidiary files all over the distribution, rather than ensuring\n>> that the configuration decisions are expressed in a central place.\n\n> I disagree here. If people want to twiddle, they can after configure. \n> People would much rather spell params to configure, rather than to edit\n> Makefiles.\n\nI agree with you, up to a point. What happens if configure makes the\nwrong choice for a given system? For a person not familiar with\nautoconf (which is most people) fixing its output is likely to be easier\nthan trying to fix the script.\n\n> ... but they MUST be required to supply the proper flags to\n> configure. We can't go down the road of allowing them to avoid the\n> 'configure' flags, and somehow enable things in Makefile.custom. There\n> is no logical way to do that, and unless we want to prevent 'configure'\n> from doing the things it does so well, we will have to live with that\n> limitation.\n\nI certainly agree we should use configure for what it's designed to do,\nnamely make intelligent configuration settings. I'm just saying that\nit is better style to record each configuration choice in only one place\n(or at least, as few places as possible) rather than make a bunch of\ncopies in a bunch of not-easily-found files.\n\nThis has advantages even if you don't consider hand adjustments to be\nimportant. Right now, if you want to find out what configure did on a\ngiven system, you need to look at twenty different files. (Some will\ngive you no new information, but which ones?) Maintenance and support\nwill be easier if there are only a couple of files that configure\nchanges, and everything else just reads those files during build.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 17:42:31 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "> I certainly agree we should use configure for what it's designed to do,\n> namely make intelligent configuration settings. I'm just saying that\n> it is better style to record each configuration choice in only one place\n> (or at least, as few places as possible) rather than make a bunch of\n> copies in a bunch of not-easily-found files.\n\nNo. There's one file, if the configure script is right...\n\n./config.cache\n\n:)\n\nTaral\n",
"msg_date": "Tue, 27 Oct 1998 16:46:53 -0600",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "> I certainly agree we should use configure for what it's designed to do,\n> namely make intelligent configuration settings. I'm just saying that\n> it is better style to record each configuration choice in only one place\n> (or at least, as few places as possible) rather than make a bunch of\n> copies in a bunch of not-easily-found files.\n> \n> This has advantages even if you don't consider hand adjustments to be\n> important. Right now, if you want to find out what configure did on a\n> given system, you need to look at twenty different files. (Some will\n> give you no new information, but which ones?) Maintenance and support\n> will be easier if there are only a couple of files that configure\n> changes, and everything else just reads those files during build.\n\nBut your Makefiles become little 'sed' monsters, making it _more_\nconfusing for people to edit the Makefiles.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 18:14:21 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "> But your Makefiles become little 'sed' monsters, making it _more_\n> confusing for people to edit the Makefiles.\n\nhmm... It's such a shame we can't insist on gmake.\n\nTaral\n",
"msg_date": "Tue, 27 Oct 1998 17:23:12 -0600",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > But your Makefiles become little 'sed' monsters, making it _more_\n> > confusing for people to edit the Makefiles.\n> \n> hmm... It's such a shame we can't insist on gmake.\n\nWe do insist on gmake. Can't compile without it.\n\nLike the term \"sed monsters\"?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 18:24:45 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "On Tue, 27 Oct 1998, Bruce Momjian wrote:\n\n> Look in Makefile.global. It is huge, and very hard to understand, even\n> for veteran developers. I agree people should be able to twiddle with\n> Makefile.custom, but they MUST be required to supply the proper flags to\n> configure. We can't go down the road of allowing them to avoid the\n> 'configure' flags, and somehow enable things in Makefile.custom. There\n> is no logical way to do that, and unless we want to prevent 'configure'\n> from doing the things it does so well, we will have to live with that\n> limitation.\n> \n> I have my configure command and flags in a script and use run it as part\n> of the cvs update I do to keep my sources current.\n\n\tMust agree here...that was what screwed me on the --with-perl :)\nI had that set to, and tried to do a 'make install' as myself, which\n*should* have worked, except that it tried to install perl stuff too...\n\n\tI've never used Makefile.custom myself, as much because I never\nremember it is there as that its never been required (configure *should*\nhandle all that)...\n\n\tIn fact...other then those that are \"veterans\", how many ppl out\nthere even know it exists? I think its more a legacy feature then\nanything...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 21:57:22 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem, redux (was Re: TCL installation\n\ttroubles)"
},
{
"msg_contents": "On Tue, 27 Oct 1998, Tom Lane wrote:\n\n> Bruce Momjian <[email protected]> writes:\n> >> Namely: I think we have gotten much too willing to use configure to\n> >> rewrite subsidiary files all over the distribution, rather than ensuring\n> >> that the configuration decisions are expressed in a central place.\n> \n> > I disagree here. If people want to twiddle, they can after configure. \n> > People would much rather spell params to configure, rather than to edit\n> > Makefiles.\n> \n> I agree with you, up to a point. What happens if configure makes the\n> wrong choice for a given system? For a person not familiar with\n> autoconf (which is most people) fixing its output is likely to be easier\n> than trying to fix the script.\n\n\tHuh? Can you give an example of a Makefile that configure\nmodifies that could pose a problem?\n\n\tOff the top of my head, the only stuff that configure modifies is\nstuff like backend/port/Makefile, to add in OS-missing functions...for the\nmost part, that should be *all* that configure modifies...and, ya, we\ncould move that into Makefile.global...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 22:00:06 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem, redux (was Re: TCL installation\n\ttroubles)"
},
{
"msg_contents": "On Tue, 27 Oct 1998, Taral wrote:\n\n> > But your Makefiles become little 'sed' monsters, making it _more_\n> > confusing for people to edit the Makefiles.\n> \n> hmm... It's such a shame we can't insist on gmake.\n\n\tHuh? We do insist on gmake:\n\n> /usr/bin/make\nYou must use GNU make to use Postgres. It may be installed\non your system with the name 'gmake'.\n\nNOTE: If you are sure that you are using GNU make and you are\n still getting this message, you may simply need to run\n the configure program.\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 22:01:08 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Configure problem, redux (was Re: TCL installation\n\ttroubles)"
},
{
"msg_contents": "\nJust curious, but exactly *what* problem are we trying to fix here? It\nsounds to me like alot of discussion is going into a solution to a problem\nthat I don't know if exists or not, since I don't recall anyone stating a\nproblem...\n\nIf configure is making 'wrong guesses', then we should be able to tighten\nup the tests so that they are correct...\n\nFrankly, and, again, I'm basing this off of memory, I can't think of\nanywhere in the 'configure hierarchy' where configure changes anything\nmajor. Most of the 'changes' are to Makefile.global and config.h (and,\nhey, I may not be the one that put the configure seed in, but am I the one\nthat cursed over a good portion of the tests and extending it)...\n\nSo, can someone state *what* the problem is?\n\nOn Tue, 27 Oct 1998, Bruce Momjian wrote:\n\n> > The problem here is that you are duplicating the normal configure\n> > processing in every Makefile that needs it. This will get old very\n> > fast, and hard to maintain. configure does this already, and\n> > automatically, in one place. Yes, you must re-run configure, and you do\n> > loose your changes, but pulling all the stuff into every Makefile seems\n> > worse.\n> > \n> > We are only doing this for the *.sql files because they don't have\n> > within them some ability to do argument expansion. This is definitely\n> > not good for all the Makefiles or source code, both of which can get\n> > what they need in a single place (Makefile.global or config.h). I am\n> > not advocating changing that. In fact, I am agreeing with the idea of\n> > becoming more centralized for those parts of the code that can handle\n> > it, i.e., Makefiles and C source.\n> \n> Oh. :-)\n> \n> I will go back to looking the other way.\n> \n> \n> -- \n> Bruce Momjian | http://www.op.net/~candle\n> [email protected] | (610) 853-3000\n> + If your life is a hard drive, | 830 Blythe Avenue\n> + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 22:10:06 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Configure problem, redux (was Re: TCL installation\n\ttroubles)"
},
{
"msg_contents": "> So, can someone state *what* the problem is?\n\nAFAICT, there's not a real problem here... Someone was just mentioning that\nit isn't easy to override the configure defaults... they appear all OVER the\nplace. (every *.in file uses them)\n\nThe original problem was that the CPPSTDIN code I wrote for configure.in\ndoesn't detect the case where neither $(CPP) nor $(CPP) - work for cpp from\nstdin... we have a possible fix (try /bin/cpp if both fail) which might go\nin...\n\nHowever, I still think that anyone who wants to override configure can\neither modify config.cache or config.status... probably the latter in most\ncases.\n\nTaral\n\n",
"msg_date": "Tue, 27 Oct 1998 20:22:06 -0600",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Re: Configure problem,\n\tredux (was Re: TCL installationtroubles)"
},
{
"msg_contents": "\"Thomas G. Lockhart\" wrote:\n> All good points. I had the USE_TCL in my Makefile.custom because for the\n> last few months/years that was the only way to get anything about tcl\n> touched by the installation afaik. I'll take it out.\n> \n> But I'm pretty sure that doesn't explain all the breakage. Will continue\n> testing a bit (I *really* need to get back to the docs!), but the first\n> problem I saw was due to a missing file which was not built\n> automatically, and if you don't do a clean install you won't see the\n> problem again. That's why I could work with \"cvs update -Pd\" for weeks\n> and not see the breakage introduced, because by that point a\n> Makefile.tcldefs already existed.\n> \n> > It's awfully late to be fixing this stuff for 6.4, unless you want to\n> > slip the release date again. But I suggest revisiting it for 6.4.1,\n> > and trying to consolidate configuration decisions into as few files\n> > as possible.\n> \n> I would strongly suggest that we fix it now, with small incremental\n> changes to make it work as currently designed. Releasing it broken\n> doesn't do much good.\n> \n> So, I'm not sure I understand what the current design is really supposed\n> to do, but istm that we could do a conditional include of\n> Makefile.tcldefs, and have Makefile.tcldefs be a prerequisite for\n> $(PGMS). Like this:\n> \n> ifneq ($(wildcard Makefile.tclsh), )\n> include Makefile.tclsh\n> endif\n> ...\n> all: Makefile.tclsh\n> $(MAKE) $(PGMS)\n> ...\n> \n> Any redesign after v6.4 is released should probably wait for v6.5, since\n> a v6.4.1 release won't get adequate testing. We managed to break the\n> Linux port of Postgres for v6.3.1 for similar reasons.\n> \n> - Tom\n\nThe breakage with the missing file occured when you defined USE_TCL in \nMakefile.custom. That caused the make file (probably in bin/plpgtcl) to try \nto include Makefile.[tcl|tk]defs. Since that file did not exist, it tried to \nbuild it by executing MkMakefile.[tcl|tk]defs.sh, which didn't exist because \nconfigure did not build it since it did not know to build the tcl/tk support \nfiles.\n\nNow for my 2 cents worth in the configure debate.\n\nWhy use configure? Why not have a makefile and config.h file that would be \nedited by hand to configure and build postgresql?\n\nPostgres\bSQL is a large system with many build options. Do you want TCL \nsupport? TK support? perl support? C++ libraries? Using autoconf and \nconfigure makes correctly building PostgreSQL for many differenct \nOS/architecture easier. If configure is going to be used, then we should use \nit to configure the system and not make it easy or desirable to circumvent \nconfigure. This mean discouraging editing of make files by hand, because \nthere is no way to know what side effects and problems will occur if hand \nediting is performed. To this end, I believe that configure should only \nsubsitute fully expanded values. If you want to change the prefix directory \n(or any other option that is handled by a configure command line option, then \nre-run configure. That is what it is there for. If configure won't create a \nbuildable version of postgresql on your system, then fix configure and send in \nthe patches.\n\nOf course, we are now entering into areas of philosophical differences which should make for interesting discussions :-)\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Thu, 29 Oct 1998 00:19:14 -0500",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Configure problem, redux (was Re: TCL installation troubles) "
}
] |
[
{
"msg_contents": "> YES! That's exactly the problem! Relinked with -Bsymbolic (which\n> necessitates -lc -lm, as well) and all's fine. Thanks.\n> > In testing on my machine the linking works correctly...\n> Yup. It doesn't use -Bsymbolic by default.\n\nDarn. It was apparently broken when the main Postgres distribution went\nto this new separate Makefile.shlib. Will look at it, and it should be\nfixed up in time for the v6.4 release. In the meantime your workaround\nis the right thing to do, including the \"-lc -lm\" flags.\n\nSorry for the pain, but it took me two weeks of poking to figure out\nthis \"misdirected linking problem\", so I'm just spreading the joy around\na little bit ;)\n\n - Thomas\n\nTom, I'm thinking of putting a \"LDFLAGS_ODBC\" into the port-specific\nMakefile, to keep this goofy set of flags separate from other shared\nlibraries. They are necessary because the iodbc driver has _exactly_ the\nsame entry point names as the psqlodbc driver, and the linker\npreferentially chooses the iodbc ones when resolving links (the setup is\nthat the iodbc generic driver dynamically links in the psqlodbc\ndb-specific one when connecting to a database). So you need to force\nthem beforehand with this flags, at least on Linux.\n",
"msg_date": "Tue, 27 Oct 1998 15:23:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: AWL: PostgreSQL ODBC."
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> Tom, I'm thinking of putting a \"LDFLAGS_ODBC\" into the port-specific\n> Makefile, to keep this goofy set of flags separate from other shared\n> libraries. They are necessary because the iodbc driver has _exactly_ the\n> same entry point names as the psqlodbc driver,\n\nYipes. I suppose it's far too late to suggest that that convention is\nguaranteed to create porting problems?\n\n> and the linker\n> preferentially chooses the iodbc ones when resolving links (the setup is\n> that the iodbc generic driver dynamically links in the psqlodbc\n> db-specific one when connecting to a database). So you need to force\n> them beforehand with this flags, at least on Linux.\n\nI suspect that some such hackery will be necessary on almost every\nplatform. Maybe you should put LDFLAGS_ODBC into the template files\ninstead of the port makefiles? I've never been real clear on why\nwe have both sets of mechanism...\n\nAnyway, I'd suggest making all of the LDFLAGS_SL assignments in\nMakefile.shlib be += rather than :=, and then the Makefile for the\nODBC driver can add on the LDFLAGS_ODBC from wherever we decide to\nget them. Are there any other hacks besides setting more flags that\nwe are likely to need?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 11:49:07 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: AWL: PostgreSQL ODBC. "
},
{
"msg_contents": "> > Tom, I'm thinking of putting a \"LDFLAGS_ODBC\" into the port-specific\n> > Makefile,\n> I suspect that some such hackery will be necessary on almost every\n> platform. Maybe you should put LDFLAGS_ODBC into the template files\n> instead of the port makefiles?\n\nYes, that's what I meant...\n\n> Anyway, I'd suggest making all of the LDFLAGS_SL assignments in\n> Makefile.shlib be += rather than :=, and then the Makefile for the\n> ODBC driver can add on the LDFLAGS_ODBC from wherever we decide to\n> get them. Are there any other hacks besides setting more flags that\n> we are likely to need?\n\nActually, I don't know why the Makefile.shlib needs anything\nplatform-specific other than perhaps helping with the naming conventions\nfor libraries. If the LDFLAGS_SL was in the platform template then\nMakefile.shlib wouldn't need anything in that regard, right?\n\n> Are there any other hacks besides setting more flags that\n> we are likely to need?\n\nDon't know. That's why they are hacks :)\n",
"msg_date": "Tue, 27 Oct 1998 17:20:48 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: AWL: PostgreSQL ODBC."
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> Actually, I don't know why the Makefile.shlib needs anything\n> platform-specific other than perhaps helping with the naming conventions\n> for libraries. If the LDFLAGS_SL was in the platform template then\n> Makefile.shlib wouldn't need anything in that regard, right?\n\nYou're right, much of that stuff could be shoved out into the\nMakefile.port files. Maybe all of it --- does anyone know GNU Make's\nsubstitution rules well enough to know whether\n\nshlib := lib$(NAME)$(DLSUFFIX).$(SO_MAJOR_VERSION).$(SO_MINOR_VERSION)\n\ncan appear *before* NAME and so forth get defined? If so, the\nlibrary naming rule could be defined by Makefile.port too. I was\nbeing conservative when I wrote Makefile.shlib, but I'm not sure\nthat the order of declarations it recommends is really essential.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 13:48:56 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: AWL: PostgreSQL ODBC. "
}
] |
[
{
"msg_contents": "somehow current CVS configure does not detect that $CC -E is not working on AIX\nand therefore configures xlc -E as preprocessor. It should be /lib/cpp\n\n???\nAndreas\n\n",
"msg_date": "Tue, 27 Oct 1998 17:49:54 +0100",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "configure and cpp detection"
},
{
"msg_contents": "Andreas Zeugswetter <[email protected]> writes:\n> somehow current CVS configure does not detect that $CC -E is not\n> working on AIX and therefore configures xlc -E as preprocessor. It\n> should be /lib/cpp\n\nHmm. We just use the standard Autoconf AC_PROG_CPP macro, and there\nare comments in the Autoconf code that show it's been tested on AIX.\n\nOr are you only complaining about the case of invoking cpp to process\nstdin? I changed the configure test for that on Sunday; maybe I\nbroke something. Or maybe AC_PROG_CPP is selecting $CC -E for CPP\nbecause that works OK for reading from a file, but it can't be made\nto work for reading stdin?\n\nWhat does the config.log file show for the cpp tests (around line 1200\nin configure)?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 12:36:43 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] configure and cpp detection "
}
] |
[
{
"msg_contents": "Needs continuation character for portability:\n\n*** s_lock.h.ori\tTue Oct 13 14:21:55 1998\n--- s_lock.h\tTue Oct 27 18:03:57 \n1998\n***************\n*** 175,181 ****\n tas(volatile slock_t *lock)\n {\n \n register _res;\n! __asm__(\"sbitb 0, %0\n \tsfsd %1\"\n \t: \"=m\"(*lock), \n\"=r\"(_res));\n return (int) _res; \n--- 175,181 ----\n tas(volatile \nslock_t *lock)\n {\n register _res;\n! __asm__(\"sbitb 0, %0 \\\n \tsfsd \n%1\"\n \t: \"=m\"(*lock), \"=r\"(_res));\n return (int) _res; \n\n",
"msg_date": "Tue, 27 Oct 1998 18:09:07 +0100",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "s_lock.h line 178"
},
{
"msg_contents": "Applied by hand. Moved %1 up to the proper line.\n\n\n> Needs continuation character for portability:\n> slock_t *lock)\n> {\n> register _res;\n> ! __asm__(\"sbitb 0, %0 \\\n> \tsfsd \n> %1\"\n> \t: \"=m\"(*lock), \"=r\"(_res));\n> return (int) _res; \n> \n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 12:46:36 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] s_lock.h line 178"
}
] |
[
{
"msg_contents": "> found an email from you, from Tue, 12 May 1998, where you describe the\n> problems I am having with regards to casting from money to other \n> types.\n> \"So there is some funny interaction on the casting, the same as you \n> found in v6.3.2 (and presumably forever), which I will look into...\"\n> \n> But these problems still exist in 6.4BETA2... Is this something that \n> will be fixed?\n> \n> To summarize the problem in case it isn't fresh in your mind: casting \n> a money column in a table to float8 results in gibberish. Casting a\n> constant seems to work fine. Casts to other types than float8 seem\n> equally broken in the table case.\n\nHow are you doing the casting? I'm having trouble figuring out what\nsyntax you are using to get gibberish...\n\n - Tom\n\npostgres=> create table mm (m money);\nCREATE\npostgres=> insert into mm values ('1');\nINSERT 18473 1\npostgres=> insert into mm values ('2.50');\nINSERT 18474 1\npostgres=> select * from mm;\nm\n-----\n$1.00\n$2.50\n(2 rows)\n\npostgres=> select float8(m) from mm;\nERROR: No such function 'float8' with the specified attributes\npostgres=> select m * 1.0 from mm;\n?column?\n--------\n$1.00\n$2.50\n(2 rows)\n\npostgres=> select m::float8 from mm;\nERROR: No such function 'float8' with the specified attributes\n",
"msg_date": "Tue, 27 Oct 1998 17:15:05 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: casting money type to float"
}
] |
[
{
"msg_contents": ">> somehow current CVS configure does not detect that $CC -E is not\n>> working on AIX and therefore configures xlc -E as preprocessor. It\n>> should be /lib/cpp\n>\n>Hmm. We just use the standard Autoconf AC_PROG_CPP macro, and there\n>are comments in the Autoconf code that show it's been tested on AIX.\n>\n>Or are you only complaining about the case of invoking cpp to process\n>stdin? I changed the configure test for that on Sunday; maybe I\n>broke something. Or maybe AC_PROG_CPP is selecting $CC -E for CPP\n>because that works OK for reading from a file, but it can't be made\n>to work for reading stdin?\n\nYes, it is only not working for the stdin case, which is first needed for Gen_fmgrtab.sh\nto create the F_OIDEQ stuff in fmgr.h\n\n> What does the config.log file show for the cpp tests (around line 1200\n> in configure)?\n\nThe line in config.log sais:\nconfigure:1241: checking how to use cpp with stdin\nxlc: 1501-221 cannot exec program less -eM - No such file or directory\n\nif I unset PAGER the log sais nothing:\nconfigure:1241: checking how to use cpp with stdin\n\n\nSorry for the confusion\nAndreas\n\n\n",
"msg_date": "Tue, 27 Oct 1998 19:12:52 +0100",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] configure and cpp detection "
},
{
"msg_contents": "Andreas Zeugswetter <[email protected]> writes:\n> Yes, it is only not working for the stdin case, which is first needed\n> for Gen_fmgrtab.sh to create the F_OIDEQ stuff in fmgr.h\n\n> The line in config.log sais:\n> configure:1241: checking how to use cpp with stdin\n> xlc: 1501-221 cannot exec program less -eM - No such file or directory\n\nWow. $(CC) tries to invoke the pager on your platform? How odd.\n\nI don't think that what I changed could have broken this --- I just\nchanged the test that decides whether configure will use \"$CPP -\" or\nplain \"$CPP\" for cpp-from-stdin. $CPP is still getting set by the same\nold AC_PROG_CPP macro. Could this have been broken before Sunday?\nHow long ago did you last run a configure and build?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 17:04:33 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] configure and cpp detection "
},
{
"msg_contents": "> I don't think that what I changed could have broken this --- I just\n> changed the test that decides whether configure will use \"$CPP -\" or\n> plain \"$CPP\" for cpp-from-stdin. $CPP is still getting set by the \n> same old AC_PROG_CPP macro. Could this have been broken before \n> Sunday? How long ago did you last run a configure and build?\n\nIt isn't your fault, Tom. *I* helped with the breakage by changing the\nMakefiles to actually use the results of an existing test in configure.\nBefore, several makefiles had references to cpp hardcoded in, and afaik\nnowhere were the configure results used. That broke when I tried\ninstalling egcs on my development system, which seemed to be missing an\nexplicit cpp.\n\nThat was maybe a month ago.\n\n - Thomas\n",
"msg_date": "Tue, 27 Oct 1998 23:03:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] configure and cpp detection"
}
] |
[
{
"msg_contents": "There is a new Postscript graphics file in doc/src/graphics/catalogs.ps.\nI assume that the comment in the log:\n\nrevision 1.1\ndate: 1998/08/11 02:32:00; author: momjian; state: Exp;\nNew findoidjoins examines oid columns to find join relationships.\n\nmay mean that something in that is different than for the original file\ncatalogs.gif. Is the source available in something other than Postscript\nor gif? I had checked in .gif files to allow building the html easily,\nand checked in the .ag (Applix graphics) files as sources I had\nconverted from the original gif files, expecting to move to another\nformat at some point. Hannu and I had concluded (last spring?) that CGM\nformat might be the most portable, but we didn't actually commit to\nchanging to that for the sources.\n\nAnyway, let me know what's available; I'm past due on starting the\nhardcopy docs integration...\n\n - Tom\n",
"msg_date": "Tue, 27 Oct 1998 20:09:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "new graphics file?"
},
{
"msg_contents": "> There is a new Postscript graphics file in doc/src/graphics/catalogs.ps.\n> I assume that the comment in the log:\n> \n> revision 1.1\n> date: 1998/08/11 02:32:00; author: momjian; state: Exp;\n> New findoidjoins examines oid columns to find join relationships.\n> \n> may mean that something in that is different than for the original file\n> catalogs.gif. Is the source available in something other than Postscript\n> or gif? I had checked in .gif files to allow building the html easily,\n> and checked in the .ag (Applix graphics) files as sources I had\n> converted from the original gif files, expecting to move to another\n> format at some point. Hannu and I had concluded (last spring?) that CGM\n> format might be the most portable, but we didn't actually commit to\n> changing to that for the sources.\n> \n> Anyway, let me know what's available; I'm past due on starting the\n> hardcopy docs integration...\n\nThe .ps file is from the 1.* release. The current flowchart gif is\nunreadable, while this one is quite nice. At one point, I saw a\nuntility that would pull diagrams out of a postscript file and into\nxfig, but I lost name of it somehow.\n\nI just left it there so we would have a readable copy we could use\nsomeday. It would be nice if we could get this loaded into xfig, or\nclick-able html. I have thought about it, and findoidjions does most of\nthe work. I have toyed with some html, and it looked OK, but I was not\n100% satisfied. I was looking for some multi-level display, that could\nshow tables, and you could click into them.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 15:58:12 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] new graphics file?"
}
] |
[
{
"msg_contents": "Someone checked in code that assumes vsnprintf() exists.\n\nShould be cool, we have an emulation right? Well, no.\nconfigure sets up to build port/snprintf.c if snprintf()\ndoesn't exist. HPUX, for one, has that but not vsnprintf.\n\nI guess the right fix is for configure to check for *both*\nsnprintf and vsnprintf --- anyone see a problem with that?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 27 Oct 1998 17:50:42 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Sigh, build is broken again"
},
{
"msg_contents": "> Someone checked in code that assumes vsnprintf() exists.\n> \n> Should be cool, we have an emulation right? Well, no.\n> configure sets up to build port/snprintf.c if snprintf()\n> doesn't exist. HPUX, for one, has that but not vsnprintf.\n> \n> I guess the right fix is for configure to check for *both*\n> snprintf and vsnprintf --- anyone see a problem with that?\n\nOops, that was me, today. Patch attached. This is not the time to be\nadding new portability calls.\n\nComments on how to handle this?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026",
"msg_date": "Tue, 27 Oct 1998 18:18:38 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Sigh, build is broken again"
},
{
"msg_contents": "On Tue, 27 Oct 1998, Bruce Momjian wrote:\n\n> > Someone checked in code that assumes vsnprintf() exists.\n> > \n> > Should be cool, we have an emulation right? Well, no.\n> > configure sets up to build port/snprintf.c if snprintf()\n> > doesn't exist. HPUX, for one, has that but not vsnprintf.\n> > \n> > I guess the right fix is for configure to check for *both*\n> > snprintf and vsnprintf --- anyone see a problem with that?\n> \n> Oops, that was me, today. Patch attached. This is not the time to be\n> adding new portability calls.\n> \n> Comments on how to handle this?\n\n\tYa...back it out until after v6.4 is released, and then we will\nhandle it then *major glare*\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 22:11:43 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Sigh, build is broken again"
},
{
"msg_contents": "On Tue, 27 Oct 1998, Tom Lane wrote:\n\n> Someone checked in code that assumes vsnprintf() exists.\n> \n> Should be cool, we have an emulation right? Well, no.\n> configure sets up to build port/snprintf.c if snprintf()\n> doesn't exist. HPUX, for one, has that but not vsnprintf.\n> \n> I guess the right fix is for configure to check for *both*\n> snprintf and vsnprintf --- anyone see a problem with that?\n\t\n\tAfter v6.4 is released, sounds great to me...before, I send guys\naround to break fingers :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 22:12:33 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Sigh, build is broken again"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n>> I guess the right fix is for configure to check for *both*\n>> snprintf and vsnprintf --- anyone see a problem with that?\n\n> Oops, that was me, today. Patch attached. This is not the time to be\n> adding new portability calls.\n\n> Comments on how to handle this?\n\nThere are two ways we can look at this:\n\n1. It's a simple oversight in configure, trivially fixable (need to\nbuild snprintf.o if *either* snprintf or vsnprintf is missing).\n\n2. We have never used vsnprintf before (if we had, the build would've\nfailed for me before). So, if we let the new trace code stand, we\nare depending on untested emulation code in snprintf.c and/or system\nlibrary routines of unproven reliability.\n\nOrdinarily I'd go for view #1 without a second thought ... but when\nwe're only four days before a major release, view #2 has to be taken\nseriously. The risk may be low, but there's not time to gain confidence\nthat vsnprintf is dependable.\n\nMarc's opinion appears to be that we should just revert the code;\nthe buffer-overrun bug is presumably of long standing, so people can\nlive with it a while longer. I think I agree with him.\n\nSo, my vote is to back out the trace.c patch, and re-apply as soon as\n6.4 is out. I'll add the necessary stuff to configure at that time too.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 11:02:25 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Sigh, build is broken again "
},
{
"msg_contents": "> Marc's opinion appears to be that we should just revert the code;\n> the buffer-overrun bug is presumably of long standing, so people can\n> live with it a while longer. I think I agree with him.\n> \n> So, my vote is to back out the trace.c patch, and re-apply as soon as\n> 6.4 is out. I'll add the necessary stuff to configure at that time too.\n\nAlready done.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 11:19:47 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Sigh, build is broken again"
},
{
"msg_contents": "Is this item resolved?\n\n> Someone checked in code that assumes vsnprintf() exists.\n> \n> Should be cool, we have an emulation right? Well, no.\n> configure sets up to build port/snprintf.c if snprintf()\n> doesn't exist. HPUX, for one, has that but not vsnprintf.\n> \n> I guess the right fix is for configure to check for *both*\n> snprintf and vsnprintf --- anyone see a problem with that?\n> \n> \t\t\tregards, tom lane\n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Fri, 11 Dec 1998 11:50:24 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Sigh, build is broken again"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> Is this item resolved?\n\n>> Someone checked in code that assumes vsnprintf() exists.\n>> \n>> Should be cool, we have an emulation right? Well, no.\n>> configure sets up to build port/snprintf.c if snprintf()\n>> doesn't exist. HPUX, for one, has that but not vsnprintf.\n>> \n>> I guess the right fix is for configure to check for *both*\n>> snprintf and vsnprintf --- anyone see a problem with that?\n\nNo, we backed out the original patch and thus made the problem go away\nfor 6.4. That's where it still stands, AFAIK.\n\nI'd suggest re-committing the original patch (something to do with\ntracing, IIRC) but to the 6.5 branch *only*. The argument about not\nhaving done enough testing applies even more strongly to 6.4.1 than\nit did for 6.4. But if we put it into 6.5 now, we should have some\ntrack record with vsnprintf before 6.5 hits the streets.\n\nAs soon as you commit the trace patch, my build will break, and then\nI'll do something about fixing configure to check for vsnprintf ;-)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Fri, 11 Dec 1998 12:02:38 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Sigh, build is broken again "
}
] |
[
{
"msg_contents": "> We do insist on gmake. Can't compile without it.\n\nThen can we use the gmake 'include' instead of the configure *.in files?\n\n> Like the term \"sed monsters\"?\n\nYes :)\n\nTaral\n",
"msg_date": "Tue, 27 Oct 1998 17:29:27 -0600",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "On Tue, 27 Oct 1998, Taral wrote:\n\n> > We do insist on gmake. Can't compile without it.\n> \n> Then can we use the gmake 'include' instead of the configure *.in files?\n\n\tUmmm...we do:\n\n> grep include */Makefile\nbackend/Makefile:include ../Makefile.global\nbin/Makefile:include ../Makefile.global\ninterfaces/Makefile:include $(SRCDIR)/Makefile.global\nlextest/Makefile:include ../Makefile.global\nman/Makefile:include ../Makefile.global\npl/Makefile:include $(SRCDIR)/Makefile.global\ntutorial/Makefile:include ../Makefile.global\nutils/Makefile:include ../Makefile.global\nutils/Makefile:include depend\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 22:01:59 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Configure problem, redux (was Re: TCL installation\n\ttroubles)"
}
] |
[
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > We do insist on gmake. Can't compile without it.\n> \n> Then can we use the gmake 'include' instead of the configure *.in files?\n> \n> > Like the term \"sed monsters\"?\n> \n> Yes :)\n\nSo you output all defines into a single file, and include that in every\nMakefile. That is interesting.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 18:36:31 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": " So you output all defines into a single file, and include that in every\n Makefile. That is interesting.\n\nI guess this isn't too clear. Here is what I am trying to advocate:\n\nconfigure output -> Makefile.global + config.h substituting as\nappropriate.\n\nall Makefiles include Makefile.global (which includes Makefile.custom\nif it exists? and other things like the shared library Makefile).\n\none of the included Makefiles has a generic rule for *.sql.in -> *.sql\nthat depends on Makefile.global also.\n\nIf a directory has *.sql.in it will automatically be converted with\nthe correct substitutions; after all, the substitution is the same in\neach case so one rule suffices (even if several variables need\nsubstituting); note that sed will not substitute for patterns not\nfound in individual *.sql.in files even if the general rule says\notherwise.\n\nSame mechanism for any other general substitutions, if necessary\n(probably not).\n\nMakefiles do not have any sed monsters, just one rule involving sed in\na generally included Makefile.sed_monsters. :) The individual\n*.sql.in files would be written in exactly the same way as if\nconfigure was doing the substitution.\n\nAll configure information is in one place (Makefile.global and\nconfig.h). All dependencies do the right thing, even if those are\nchanged post-configure.\n\nCheers,\nBrook\n",
"msg_date": "Tue, 27 Oct 1998 17:11:44 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "On Tue, 27 Oct 1998, Bruce Momjian wrote:\n\n> [Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > > We do insist on gmake. Can't compile without it.\n> > \n> > Then can we use the gmake 'include' instead of the configure *.in files?\n> > \n> > > Like the term \"sed monsters\"?\n> > \n> > Yes :)\n> \n> So you output all defines into a single file, and include that in every\n> Makefile. That is interesting.\n\n\tTechnically, we are doing that now with Makefile.global ... its\nincluded in every Makefile in the system, and has CFLAGS and such defined\nin it. It wouldn't be unreasonable to have a MISSING_OBJ= put into\nMakefile.global, and remove the Makefile.in in the backend/ports\ndirectory... *shrug*\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 27 Oct 1998 22:04:28 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem, redux (was Re: TCL installation\n\ttroubles)"
},
{
"msg_contents": "> \tTechnically, we are doing that now with Makefile.global ... its\n> included in every Makefile in the system, and has CFLAGS and such defined\n> in it. It wouldn't be unreasonable to have a MISSING_OBJ= put into\n> Makefile.global, and remove the Makefile.in in the backend/ports\n> directory... *shrug*\n\nYes. I think that kind of thing would make sense, and would allow\nMakefile.custom to over-rule it, though that is more of an affect rather\nthan a feature I would promote. configure flags should do most of the\nwork.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Tue, 27 Oct 1998 21:23:27 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Configure problem,\n\tredux (was Re: TCL installation troubles)"
},
{
"msg_contents": "Brook Milligan wrote:\n> So you output all defines into a single file, and include that in every\n> Makefile. That is interesting.\n> \n> I guess this isn't too clear. Here is what I am trying to advocate:\n> \n> configure output -> Makefile.global + config.h substituting as\n> appropriate.\n> \n> all Makefiles include Makefile.global (which includes Makefile.custom\n> if it exists? and other things like the shared library Makefile).\n> \n> one of the included Makefiles has a generic rule for *.sql.in -> *.sql\n> that depends on Makefile.global also.\n> \n> If a directory has *.sql.in it will automatically be converted with\n> the correct substitutions; after all, the substitution is the same in\n> each case so one rule suffices (even if several variables need\n> substituting); note that sed will not substitute for patterns not\n> found in individual *.sql.in files even if the general rule says\n> otherwise.\n> \n> Same mechanism for any other general substitutions, if necessary\n> (probably not).\n> \n> Makefiles do not have any sed monsters, just one rule involving sed in\n> a generally included Makefile.sed_monsters. :) The individual\n> *.sql.in files would be written in exactly the same way as if\n> configure was doing the substitution.\n> \n> All configure information is in one place (Makefile.global and\n> config.h). All dependencies do the right thing, even if those are\n> changed post-configure.\n> \n> Cheers,\n> Brook\n> \n\nThis would seem to work, although we would at least need a rule for *.sh.in -> \n*.sh for the mkMakefile.[tcl|tk]defs.sh files needed for TCL/TK support.\n\nI still have a problem with editing files to curcumvent configure. But I \ncould live with a solution as outlined above. I don't have to edit the \nmakefiles and if someone else wants to, well they roll the die and take their \nchances :-)\n\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Thu, 29 Oct 1998 01:26:35 -0500",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Configure problem, redux (was Re: TCL installation\n\ttroubles)"
}
] |
[
{
"msg_contents": "\nI was hoping someone could shed some light on the \nfollowing problem:\n\nThe subselect under 6.4 ignores indexes built on \ntables. For example, consider the following two\ntables:\n\n\\d sales;\n\nTable = sales\n\nsupplysource varchar() not null 16\nsupply varchar() not null 16\nsupplyunit varchar() not null 2\nquantity float8 not null 8\ntarget varchar() not null 16\ncostcntr varchar() not null 8\nsaletype varchar() not null 16\nsaledate datetime not null 8\n\nIndices: k_sales\n k_sales_saledate\n k_sales_supply\n k_sales_target \n\n\n\\d locations;\n\nTable = locations\nlocation varchar() not null 16\ncostcntr varchar() not null 5\nsupplypath varchar() not null 64\nformpath varchar() not null 64\nengineerpath varchar() not null 64\n\nIndices: k_locations\n k_locations_costcntr\n k_locations_location \n\nwith the following query:\n\nexplain select supply from sales where target in\n(select location from locations);\nNOTICE: QUERY PLAN:\n\nSeq Scan on sales (cost=5738.60 size=116806 width=12)\n SubPlan\n -> Seq Scan on locations (cost=7.49 size=136 width=12)\n\nEXPLAIN\n\nThis results in a full table scan on the outer table\nwhere there are 150K rows. As a result, it is the\nequivalent of a full table scan on 150K (sales) * \n100 (locations) rows = 15,000,000 row table scan.\n\nOracle 8.0.5.00 for Linux Early Adaptor Program \ncorrectly uses the index on the same two tables\nresulting on a sequential access on the subselected\ntable (locations) and an index look-up on the \nselect table (sales), according to EXPLAIN PLAN.\n\nDue to the failure of PostgreSQL 6.4 to correctly\nuse indexes with subselects, it makes subselects\npractically unusable.\n\n(NOTE: This occurs regardless of whether or not it\nis executed immediately following a VACUUM)\n\nAre there any patches available to fix this problem?\nWill this fixed in the 6.4 release?\n\nThanks for any information,\n\nMarcus Mascari\n([email protected])\n\n\n\n\n_________________________________________________________\nDO YOU YAHOO!?\nGet your free @yahoo.com address at http://mail.yahoo.com\n\n",
"msg_date": "Tue, 27 Oct 1998 17:30:44 -0800 (PST)",
"msg_from": "Marcus Mascari <[email protected]>",
"msg_from_op": true,
"msg_subject": "Please help (Someone?) - subselect not using indexes"
}
] |
[
{
"msg_contents": "> > > _ NetBSD 1.3 NS32532\n> Sorry I can't confirm or deny that this works on my machine right\n> now. I tried to upgrade it to NetBSD-current a couple of months\n> ago, and mostly got there.\n\nI don't have my heart set on v1.3, the NS32532 and the NetBSD is the\ninteresting part. If -current can build the system, then I'm happy to\nreport that version. I'll go ahead and do that, unless you say that it\nwould be erroneous to do so.\n\n> Might want to say that it \"mostly works\" with known problems in some\n> of the date code, and possibly other bugs, but still mostly usable.\n> (I don't suspect that too many people will be running this on a NS32K\n> machine anyway.)\n\nOK.\n\n - Thomas\n",
"msg_date": "Wed, 28 Oct 1998 03:15:23 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Last call?"
}
] |
[
{
"msg_contents": "Additions\n---------\nCREATE TABLE test (x text, s serial) fails if no database creation permission\nregression test all platforms\n\nSerious Items\n------------\nchange pg args for platforms that don't support argv changes\n\t(setproctitle()?, sendmail hack?)\n\nDocs\n----\ngenerate html/postscript documentation\n(User's Guide, Administrator's Guide, Programmer's Guide) (Thomas)\nmake sure all changes are documented properly\nmarkup and merge JDBC docs from Peter (Thomas, others??)\nmerge the release notes into doc/src/sgml/release.sgml (Bruce, Thomas)\ngenerate new text-format INSTALL and README from sgml sources (Thomas)\nmarkup of Jan's PL docs\n\nMinor items\n-----------\ncnf-ify still can exhaust memory, make SET KSQO more generic\npermissions on indexes: what do they do? should it be prevented?\nallow multiple generic operators in expressions without the use of parentheses\ndocument/trigger/rule so changes to pg_shadow create pg_pwd\nlarge objects orphanage\nimprove group handling\nimprove PRIMARY KEY handling\ngenerate postmaster pid file and remove flock/fcntl lock code\nadd ability to specifiy location of lock/socket files\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 00:17:54 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Open 6.4 items"
},
{
"msg_contents": "Bruce Momjian wrote:\n>\n> Additions\n> ---------\n> CREATE TABLE test (x text, s serial) fails if no database creation permission\n> regression test all platforms\n\n The patch below arranges that the sequence(s) get created\n first. Unprivileged user now can create table with serial\n columns.\n\n Regression tested.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n*** analyze.c.orig\tWed Oct 28 14:50:47 1998\n--- analyze.c\tWed Oct 28 14:59:12 1998\n***************\n*** 42,48 ****\n static Query *transformCursorStmt(ParseState *pstate, SelectStmt *stmt);\n static Query *transformCreateStmt(ParseState *pstate, CreateStmt *stmt);\n \n! List\t *extras = NIL;\n \n /*\n * parse_analyze -\n--- 42,49 ----\n static Query *transformCursorStmt(ParseState *pstate, SelectStmt *stmt);\n static Query *transformCreateStmt(ParseState *pstate, CreateStmt *stmt);\n \n! List\t *extras_before = NIL;\n! List\t *extras_after = NIL;\n \n /*\n * parse_analyze -\n***************\n*** 57,62 ****\n--- 58,64 ----\n {\n \tQueryTreeList *result;\n \tParseState *pstate;\n+ \tQuery *parsetree;\n \tint\t\t\ti = 0;\n \n \tresult = malloc(sizeof(QueryTreeList));\n***************\n*** 66,88 ****\n \twhile (pl != NIL)\n \t{\n \t\tpstate = make_parsestate(parentParseState);\n! \t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(pl));\n \t\tif (pstate->p_target_relation != NULL)\n \t\t\theap_close(pstate->p_target_relation);\n \n! \t\tif (extras != NIL)\n \t\t{\n! \t\t\tresult->len += length(extras);\n \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n! \t\t\twhile (extras != NIL)\n \t\t\t{\n! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras));\n \t\t\t\tif (pstate->p_target_relation != NULL)\n \t\t\t\t\theap_close(pstate->p_target_relation);\n! \t\t\t\textras = lnext(extras);\n \t\t\t}\n \t\t}\n! \t\textras = NIL;\n \t\tpl = lnext(pl);\n \t\tpfree(pstate);\n \t}\n--- 68,107 ----\n \twhile (pl != NIL)\n \t{\n \t\tpstate = make_parsestate(parentParseState);\n! \t\tparsetree = transformStmt(pstate, lfirst(pl));\n \t\tif (pstate->p_target_relation != NULL)\n \t\t\theap_close(pstate->p_target_relation);\n \n! \t\tif (extras_before != NIL)\n \t\t{\n! \t\t\tresult->len += length(extras_before);\n \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n! \t\t\twhile (extras_before != NIL)\n \t\t\t{\n! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras_before));\n \t\t\t\tif (pstate->p_target_relation != NULL)\n \t\t\t\t\theap_close(pstate->p_target_relation);\n! \t\t\t\textras_before = lnext(extras_before);\n \t\t\t}\n \t\t}\n! \t\textras_before = NIL;\n! \n! \t\tresult->qtrees[i++] = parsetree;\n! \n! \t\tif (extras_after != NIL)\n! \t\t{\n! \t\t\tresult->len += length(extras_after);\n! \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n! \t\t\twhile (extras_after != NIL)\n! \t\t\t{\n! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras_after));\n! \t\t\t\tif (pstate->p_target_relation != NULL)\n! \t\t\t\t\theap_close(pstate->p_target_relation);\n! \t\t\t\textras_after = lnext(extras_after);\n! \t\t\t}\n! \t\t}\n! \t\textras_after = NIL;\n! \n \t\tpl = lnext(pl);\n \t\tpfree(pstate);\n \t}\n***************\n*** 487,492 ****\n--- 506,512 ----\n \tConstraint *constraint;\n \tList\t *keys;\n \tIdent\t *key;\n+ \tList\t *blist = NIL;\n \tList\t *ilist = NIL;\n \tIndexStmt *index;\n \tIndexElem *iparam;\n***************\n*** 553,559 ****\n \t\t\t\t\telog(NOTICE, \"CREATE TABLE will create implicit sequence %s for SERIAL column %s.%s\",\n \t\t\t\t\t sequence->seqname, stmt->relname, column->colname);\n \n! \t\t\t\t\tilist = lcons(sequence, NIL);\n \t\t\t\t}\n \n \t\t\t\tif (column->constraints != NIL)\n--- 573,579 ----\n \t\t\t\t\telog(NOTICE, \"CREATE TABLE will create implicit sequence %s for SERIAL column %s.%s\",\n \t\t\t\t\t sequence->seqname, stmt->relname, column->colname);\n \n! \t\t\t\t\tblist = lcons(sequence, NIL);\n \t\t\t\t}\n \n \t\t\t\tif (column->constraints != NIL)\n***************\n*** 745,751 ****\n \t}\n \n \tq->utilityStmt = (Node *) stmt;\n! \textras = ilist;\n \n \treturn q;\n }\n--- 765,772 ----\n \t}\n \n \tq->utilityStmt = (Node *) stmt;\n! \textras_before = blist;\n! \textras_after = ilist;\n \n \treturn q;\n }\n",
"msg_date": "Wed, 28 Oct 1998 15:19:03 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Wed, 28 Oct 1998, Jan Wieck wrote:\n\n> Bruce Momjian wrote:\n> >\n> > Additions\n> > ---------\n> > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > regression test all platforms\n> \n> The patch below arranges that the sequence(s) get created\n> first. Unprivileged user now can create table with serial\n> columns.\n\n\tIs this safe to leave until post-v6.4 released? \n> \n> Regression tested.\n> \n> \n> Jan\n> \n> --\n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n> \n> \n> *** analyze.c.orig\tWed Oct 28 14:50:47 1998\n> --- analyze.c\tWed Oct 28 14:59:12 1998\n> ***************\n> *** 42,48 ****\n> static Query *transformCursorStmt(ParseState *pstate, SelectStmt *stmt);\n> static Query *transformCreateStmt(ParseState *pstate, CreateStmt *stmt);\n> \n> ! List\t *extras = NIL;\n> \n> /*\n> * parse_analyze -\n> --- 42,49 ----\n> static Query *transformCursorStmt(ParseState *pstate, SelectStmt *stmt);\n> static Query *transformCreateStmt(ParseState *pstate, CreateStmt *stmt);\n> \n> ! List\t *extras_before = NIL;\n> ! List\t *extras_after = NIL;\n> \n> /*\n> * parse_analyze -\n> ***************\n> *** 57,62 ****\n> --- 58,64 ----\n> {\n> \tQueryTreeList *result;\n> \tParseState *pstate;\n> + \tQuery *parsetree;\n> \tint\t\t\ti = 0;\n> \n> \tresult = malloc(sizeof(QueryTreeList));\n> ***************\n> *** 66,88 ****\n> \twhile (pl != NIL)\n> \t{\n> \t\tpstate = make_parsestate(parentParseState);\n> ! \t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(pl));\n> \t\tif (pstate->p_target_relation != NULL)\n> \t\t\theap_close(pstate->p_target_relation);\n> \n> ! \t\tif (extras != NIL)\n> \t\t{\n> ! \t\t\tresult->len += length(extras);\n> \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n> ! \t\t\twhile (extras != NIL)\n> \t\t\t{\n> ! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras));\n> \t\t\t\tif (pstate->p_target_relation != NULL)\n> \t\t\t\t\theap_close(pstate->p_target_relation);\n> ! \t\t\t\textras = lnext(extras);\n> \t\t\t}\n> \t\t}\n> ! \t\textras = NIL;\n> \t\tpl = lnext(pl);\n> \t\tpfree(pstate);\n> \t}\n> --- 68,107 ----\n> \twhile (pl != NIL)\n> \t{\n> \t\tpstate = make_parsestate(parentParseState);\n> ! \t\tparsetree = transformStmt(pstate, lfirst(pl));\n> \t\tif (pstate->p_target_relation != NULL)\n> \t\t\theap_close(pstate->p_target_relation);\n> \n> ! \t\tif (extras_before != NIL)\n> \t\t{\n> ! \t\t\tresult->len += length(extras_before);\n> \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n> ! \t\t\twhile (extras_before != NIL)\n> \t\t\t{\n> ! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras_before));\n> \t\t\t\tif (pstate->p_target_relation != NULL)\n> \t\t\t\t\theap_close(pstate->p_target_relation);\n> ! \t\t\t\textras_before = lnext(extras_before);\n> \t\t\t}\n> \t\t}\n> ! \t\textras_before = NIL;\n> ! \n> ! \t\tresult->qtrees[i++] = parsetree;\n> ! \n> ! \t\tif (extras_after != NIL)\n> ! \t\t{\n> ! \t\t\tresult->len += length(extras_after);\n> ! \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n> ! \t\t\twhile (extras_after != NIL)\n> ! \t\t\t{\n> ! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras_after));\n> ! \t\t\t\tif (pstate->p_target_relation != NULL)\n> ! \t\t\t\t\theap_close(pstate->p_target_relation);\n> ! \t\t\t\textras_after = lnext(extras_after);\n> ! \t\t\t}\n> ! \t\t}\n> ! \t\textras_after = NIL;\n> ! \n> \t\tpl = lnext(pl);\n> \t\tpfree(pstate);\n> \t}\n> ***************\n> *** 487,492 ****\n> --- 506,512 ----\n> \tConstraint *constraint;\n> \tList\t *keys;\n> \tIdent\t *key;\n> + \tList\t *blist = NIL;\n> \tList\t *ilist = NIL;\n> \tIndexStmt *index;\n> \tIndexElem *iparam;\n> ***************\n> *** 553,559 ****\n> \t\t\t\t\telog(NOTICE, \"CREATE TABLE will create implicit sequence %s for SERIAL column %s.%s\",\n> \t\t\t\t\t sequence->seqname, stmt->relname, column->colname);\n> \n> ! \t\t\t\t\tilist = lcons(sequence, NIL);\n> \t\t\t\t}\n> \n> \t\t\t\tif (column->constraints != NIL)\n> --- 573,579 ----\n> \t\t\t\t\telog(NOTICE, \"CREATE TABLE will create implicit sequence %s for SERIAL column %s.%s\",\n> \t\t\t\t\t sequence->seqname, stmt->relname, column->colname);\n> \n> ! \t\t\t\t\tblist = lcons(sequence, NIL);\n> \t\t\t\t}\n> \n> \t\t\t\tif (column->constraints != NIL)\n> ***************\n> *** 745,751 ****\n> \t}\n> \n> \tq->utilityStmt = (Node *) stmt;\n> ! \textras = ilist;\n> \n> \treturn q;\n> }\n> --- 765,772 ----\n> \t}\n> \n> \tq->utilityStmt = (Node *) stmt;\n> ! \textras_before = blist;\n> ! \textras_after = ilist;\n> \n> \treturn q;\n> }\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 28 Oct 1998 11:14:55 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Thanks. Applied.\n\n> Bruce Momjian wrote:\n> >\n> > Additions\n> > ---------\n> > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > regression test all platforms\n> \n> The patch below arranges that the sequence(s) get created\n> first. Unprivileged user now can create table with serial\n> columns.\n> \n> Regression tested.\n> \n> \n> Jan\n> \n> --\n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n> \n> \n> *** analyze.c.orig\tWed Oct 28 14:50:47 1998\n> --- analyze.c\tWed Oct 28 14:59:12 1998\n> ***************\n> *** 42,48 ****\n> static Query *transformCursorStmt(ParseState *pstate, SelectStmt *stmt);\n> static Query *transformCreateStmt(ParseState *pstate, CreateStmt *stmt);\n> \n> ! List\t *extras = NIL;\n> \n> /*\n> * parse_analyze -\n> --- 42,49 ----\n> static Query *transformCursorStmt(ParseState *pstate, SelectStmt *stmt);\n> static Query *transformCreateStmt(ParseState *pstate, CreateStmt *stmt);\n> \n> ! List\t *extras_before = NIL;\n> ! List\t *extras_after = NIL;\n> \n> /*\n> * parse_analyze -\n> ***************\n> *** 57,62 ****\n> --- 58,64 ----\n> {\n> \tQueryTreeList *result;\n> \tParseState *pstate;\n> + \tQuery *parsetree;\n> \tint\t\t\ti = 0;\n> \n> \tresult = malloc(sizeof(QueryTreeList));\n> ***************\n> *** 66,88 ****\n> \twhile (pl != NIL)\n> \t{\n> \t\tpstate = make_parsestate(parentParseState);\n> ! \t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(pl));\n> \t\tif (pstate->p_target_relation != NULL)\n> \t\t\theap_close(pstate->p_target_relation);\n> \n> ! \t\tif (extras != NIL)\n> \t\t{\n> ! \t\t\tresult->len += length(extras);\n> \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n> ! \t\t\twhile (extras != NIL)\n> \t\t\t{\n> ! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras));\n> \t\t\t\tif (pstate->p_target_relation != NULL)\n> \t\t\t\t\theap_close(pstate->p_target_relation);\n> ! \t\t\t\textras = lnext(extras);\n> \t\t\t}\n> \t\t}\n> ! \t\textras = NIL;\n> \t\tpl = lnext(pl);\n> \t\tpfree(pstate);\n> \t}\n> --- 68,107 ----\n> \twhile (pl != NIL)\n> \t{\n> \t\tpstate = make_parsestate(parentParseState);\n> ! \t\tparsetree = transformStmt(pstate, lfirst(pl));\n> \t\tif (pstate->p_target_relation != NULL)\n> \t\t\theap_close(pstate->p_target_relation);\n> \n> ! \t\tif (extras_before != NIL)\n> \t\t{\n> ! \t\t\tresult->len += length(extras_before);\n> \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n> ! \t\t\twhile (extras_before != NIL)\n> \t\t\t{\n> ! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras_before));\n> \t\t\t\tif (pstate->p_target_relation != NULL)\n> \t\t\t\t\theap_close(pstate->p_target_relation);\n> ! \t\t\t\textras_before = lnext(extras_before);\n> \t\t\t}\n> \t\t}\n> ! \t\textras_before = NIL;\n> ! \n> ! \t\tresult->qtrees[i++] = parsetree;\n> ! \n> ! \t\tif (extras_after != NIL)\n> ! \t\t{\n> ! \t\t\tresult->len += length(extras_after);\n> ! \t\t\tresult->qtrees = (Query **) realloc(result->qtrees, result->len * sizeof(Query *));\n> ! \t\t\twhile (extras_after != NIL)\n> ! \t\t\t{\n> ! \t\t\t\tresult->qtrees[i++] = transformStmt(pstate, lfirst(extras_after));\n> ! \t\t\t\tif (pstate->p_target_relation != NULL)\n> ! \t\t\t\t\theap_close(pstate->p_target_relation);\n> ! \t\t\t\textras_after = lnext(extras_after);\n> ! \t\t\t}\n> ! \t\t}\n> ! \t\textras_after = NIL;\n> ! \n> \t\tpl = lnext(pl);\n> \t\tpfree(pstate);\n> \t}\n> ***************\n> *** 487,492 ****\n> --- 506,512 ----\n> \tConstraint *constraint;\n> \tList\t *keys;\n> \tIdent\t *key;\n> + \tList\t *blist = NIL;\n> \tList\t *ilist = NIL;\n> \tIndexStmt *index;\n> \tIndexElem *iparam;\n> ***************\n> *** 553,559 ****\n> \t\t\t\t\telog(NOTICE, \"CREATE TABLE will create implicit sequence %s for SERIAL column %s.%s\",\n> \t\t\t\t\t sequence->seqname, stmt->relname, column->colname);\n> \n> ! \t\t\t\t\tilist = lcons(sequence, NIL);\n> \t\t\t\t}\n> \n> \t\t\t\tif (column->constraints != NIL)\n> --- 573,579 ----\n> \t\t\t\t\telog(NOTICE, \"CREATE TABLE will create implicit sequence %s for SERIAL column %s.%s\",\n> \t\t\t\t\t sequence->seqname, stmt->relname, column->colname);\n> \n> ! \t\t\t\t\tblist = lcons(sequence, NIL);\n> \t\t\t\t}\n> \n> \t\t\t\tif (column->constraints != NIL)\n> ***************\n> *** 745,751 ****\n> \t}\n> \n> \tq->utilityStmt = (Node *) stmt;\n> ! \textras = ilist;\n> \n> \treturn q;\n> }\n> --- 765,772 ----\n> \t}\n> \n> \tq->utilityStmt = (Node *) stmt;\n> ! \textras_before = blist;\n> ! \textras_after = ilist;\n> \n> \treturn q;\n> }\n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 11:06:15 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> Marc wrote:\n> \n> >\n> > On Wed, 28 Oct 1998, Jan Wieck wrote:\n> >\n> > > Bruce Momjian wrote:\n> > > >\n> > > > Additions\n> > > > ---------\n> > > > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > > > regression test all platforms\n> > >\n> > > The patch below arranges that the sequence(s) get created\n> > > first. Unprivileged user now can create table with serial\n> > > columns.\n> >\n> > Is this safe to leave until post-v6.4 released?\n> \n> I just did it now to let Bruce get rid of the open item. If\n> it shouldn't be in 6.4, why is it on it's open item list\n> then?\n\nIt has been on the open items. In fact, it is the only hot item, except\nfor the 'pg args' thing, which is not as hot.\n\nIt is serious breakage, in the sense that if we don't fix it, we must\nremove reference to the new SERIAL type.\n\nAs to why no one fixed it earlier, ...? Must be the same reason 'ps\nargs' is not fixed. :-)\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 11:12:21 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Marc wrote:\n\n>\n> On Wed, 28 Oct 1998, Jan Wieck wrote:\n>\n> > Bruce Momjian wrote:\n> > >\n> > > Additions\n> > > ---------\n> > > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > > regression test all platforms\n> >\n> > The patch below arranges that the sequence(s) get created\n> > first. Unprivileged user now can create table with serial\n> > columns.\n>\n> Is this safe to leave until post-v6.4 released?\n\n I just did it now to let Bruce get rid of the open item. If\n it shouldn't be in 6.4, why is it on it's open item list\n then?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Wed, 28 Oct 1998 17:16:35 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> >\n> > Thanks. Applied.\n> >\n> > > Bruce Momjian wrote:\n> > > >\n> > > > Additions\n> > > > ---------\n> > > > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > > > regression test all platforms\n> > >\n> > > The patch below arranges that the sequence(s) get created\n> > > first. Unprivileged user now can create table with serial\n> > > columns.\n> > >\n> > > Regression tested.\n> \n> So Marc's assumption that his BETA3 is what we release was\n> false. We really know how to trigger his hot buttons - eh?\n> :-)\n\nYou want even worse. I fixed a problem in the system table just now\nwith the lseg_eq <> problem. That requires an initdb for people to see\nthe fix.\n\nBut we might as well fix it now, because we can't ask for initdb after\nthe final release.\n\nCan someone check that that fix does not affect the regression test\nresults? I doubt we do lseg not-equal tests in the regression suite.\nYou will have to do an initdb to see the change. \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 11:18:50 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": ">\n> Thanks. Applied.\n>\n> > Bruce Momjian wrote:\n> > >\n> > > Additions\n> > > ---------\n> > > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > > regression test all platforms\n> >\n> > The patch below arranges that the sequence(s) get created\n> > first. Unprivileged user now can create table with serial\n> > columns.\n> >\n> > Regression tested.\n\n So Marc's assumption that his BETA3 is what we release was\n false. We really know how to trigger his hot buttons - eh?\n :-)\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Wed, 28 Oct 1998 17:31:16 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> It is serious breakage, in the sense that if we don't fix it, we must\n> remove reference to the new SERIAL type.\n> As to why no one fixed it earlier, ...? Must be the same reason 'ps\n> args' is not fixed. :-)\n\nI'm comfortable with Jan's patches, and it's my code he's mucking with\n:) For the transformation from serial to int4 with index and sequence,\nthe only time that code gets traversed is for the serial type. So it\nshouldn't have hidden side effects (yeah, sure...)\n\n'Course, I see Bruce has already committed it anyway...\n\n - Tom\n",
"msg_date": "Wed, 28 Oct 1998 16:46:18 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> So Marc's assumption that his BETA3 is what we release was\n> false. We really know how to trigger his hot buttons - eh?\n> :-)\n\nGood point. I just committed a _very_ small bit of code in the ODBC\ninterface which is non-essential, but potentially useful.\n\nAlso, he will have to rebuild a tarball one more time anyway to get the\nnew docs, which won't be finished until the last minute.\n\nJust to get his blood pressure up, I'll mention that there is a chance\nI'll ask for a couple of extra days to finish docs, but I'm hoping to\nhave them done on time :)\n\n - Tom\n",
"msg_date": "Wed, 28 Oct 1998 17:18:16 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> But we might as well fix it now, because we can't ask for initdb after\n> the final release.\n\n> Can someone check that that fix does not affect the regression test\n> results? I doubt we do lseg not-equal tests in the regression suite.\n> You will have to do an initdb to see the change. \n\nI just rebuilt and initdb'd. Regression test state is the same\nas before, except that the new inet test fails:\n\n*** expected/inet.out\tTue Oct 27 14:34:02 1998\n--- results/inet.out\tWed Oct 28 14:02:45 1998\n***************\n*** 53,66 ****\n i as inet, network(i) as \"network(inet)\" FROM INET_TBL;\n eight|cidr |network(cidr)|inet |network(inet)\n -----+------------+-------------+----------------+-------------\n! |192.168.1/24| 0.1.168.192|192.168.1.226/24| 0.1.168.192\n! |192.168.1/24| 0.1.168.192|192.168.1.226 |226.1.168.192\n! |10/8 | 0.0.0.10|10.1.2.3/8 | 0.0.0.10\n! |10.0.0.0/32 | 0.0.0.10|10.1.2.3/8 | 0.0.0.10\n! |10.1.2.3/32 | 3.2.1.10|10.1.2.3 | 3.2.1.10\n! |10.1.2/24 | 0.2.1.10|10.1.2.3/24 | 0.2.1.10\n! |10.1/16 | 0.0.1.10|10.1.2.3/16 | 0.0.1.10\n! |10/8 | 0.0.0.10|10.1.2.3/8 | 0.0.0.10\n (8 rows)\n \n QUERY: SELECT '' as eight, c as cidr, masklen(c) as \"masklen(cidr)\",\n--- 53,66 ----\n i as inet, network(i) as \"network(inet)\" FROM INET_TBL;\n eight|cidr |network(cidr)|inet |network(inet)\n -----+------------+-------------+----------------+-------------\n! |192.168.1/24| 192.168.1.0|192.168.1.226/24| 192.168.1.0\n! |192.168.1/24| 192.168.1.0|192.168.1.226 |192.168.1.226\n! |10/8 | 10.0.0.0|10.1.2.3/8 | 10.0.0.0\n! |10.0.0.0/32 | 10.0.0.0|10.1.2.3/8 | 10.0.0.0\n! |10.1.2.3/32 | 10.1.2.3|10.1.2.3 | 10.1.2.3\n! |10.1.2/24 | 10.1.2.0|10.1.2.3/24 | 10.1.2.0\n! |10.1/16 | 10.1.0.0|10.1.2.3/16 | 10.1.0.0\n! |10/8 | 10.0.0.0|10.1.2.3/8 | 10.0.0.0\n (8 rows)\n \n QUERY: SELECT '' as eight, c as cidr, masklen(c) as \"masklen(cidr)\",\n\n----------------------\n\nOffhand I would say that it's the \"expected\" file that is broken.\nShouldn't those octets be coming out in the other order?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 14:16:42 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "On Wed, 28 Oct 1998, Jan Wieck wrote:\n\n> >\n> > Thanks. Applied.\n> >\n> > > Bruce Momjian wrote:\n> > > >\n> > > > Additions\n> > > > ---------\n> > > > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > > > regression test all platforms\n> > >\n> > > The patch below arranges that the sequence(s) get created\n> > > first. Unprivileged user now can create table with serial\n> > > columns.\n> > >\n> > > Regression tested.\n> \n> So Marc's assumption that his BETA3 is what we release was\n> false. We really know how to trigger his hot buttons - eh?\n> :-)\n\n\tIf it was 'the release', it wouldn't have been named BETA :) I'll\nbuild a BETA4 tomorrow...release set for Monday...\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 28 Oct 1998 15:19:38 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
}
] |
[
{
"msg_contents": "> Except that now causes an illegal instruction fault, since the\n> assembler thinks the second instruction is some of the args to\n> the first. Will '\\n\\' work instead of '\\' as I modified the\n> patch below?\n\nYes, sorry, therefore only this will work.\n\n*** s_lock.h.ori\tTue Oct 13 14:21:55 1998\n--- s_lock.h\tTue Oct 27 18:03:57 \n1998\n***************\n*** 175,181 ****\n tas(volatile slock_t *lock)\n {\n \n register _res;\n! __asm__(\"sbitb 0, %0\n \tsfsd %1\"\n \t: \"=m\"(*lock), \n\"=r\"(_res));\n return (int) _res; \n--- 175,181 ----\n tas(volatile \nslock_t *lock)\n {\n register _res;\n! __asm__(\"sbitb 0, %0 \\n\\\n \tsfsd %1\"\n \t: \"=m\"(*lock), \"=r\"(_res));\n return (int) _res; \n\n\n",
"msg_date": "Wed, 28 Oct 1998 09:11:45 +0100",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] s_lock.h line 178 "
},
{
"msg_contents": "Applied manually. Patch was wrapped.\n\n\n> > Except that now causes an illegal instruction fault, since the\n> > assembler thinks the second instruction is some of the args to\n> > the first. Will '\\n\\' work instead of '\\' as I modified the\n> > patch below?\n> \n> Yes, sorry, therefore only this will work.\n> \n> *** s_lock.h.ori\tTue Oct 13 14:21:55 1998\n> --- s_lock.h\tTue Oct 27 18:03:57 \n> 1998\n> ***************\n> *** 175,181 ****\n> tas(volatile slock_t *lock)\n> {\n> \n> register _res;\n> ! __asm__(\"sbitb 0, %0\n> \tsfsd %1\"\n> \t: \"=m\"(*lock), \n> \"=r\"(_res));\n\n^^^^^^^^^^^^^^^^^^^^^^^^^^\n> return (int) _res; \n> --- 175,181 ----\n> tas(volatile \n> slock_t *lock)\n> {\n> register _res;\n> ! __asm__(\"sbitb 0, %0 \\n\\\n> \tsfsd %1\"\n> \t: \"=m\"(*lock), \"=r\"(_res));\n> return (int) _res; \n> \n> \n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 10:58:40 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] s_lock.h line 178"
}
] |
[
{
"msg_contents": "Hi,\n\nAIX 4.2.1 regression tests are ok :-)\n\nint2 .. failed\t\tother errmsg (ok)\nint4 .. failed\t\tother errmsg (ok)\nfloat8 .. failed\t\thas correct output / expected is wrong (ok)\ngeometry .. failed\trounding diffs (ok)\nabstime .. failed\tPDT instead of PST in some rows, 1h difference (ok)\ntinterval .. failed\t\t--\"--\nhorology .. failed\t--\"--\ninet .. failed\t\texpected/inet.out is incorrect (ok, but please remake)\nrules .. failed\t\tother sort order in select * from rtest_admin (guess ok)\n\nIt would need only the following modifikations to current CVS:\n\nremove file src/makefiles/Makefile.aix4\n\tno harm because it is not used (it does not work 100% because it won't resolve variables\n\tfrom postgres main executable, therefore the postgres.imp is still needed) .\n\talternatively I could try to do a new one\nmodify configure test for cpp stdin\n\tcurrently does xlc -E and fails to notice, that it does not work\nhand made libplpgsql.so\n\tthe command was:\n\t\tld -H512 -bM:SRE -bI:/usr/postgres/lib/postgres.imp -bexpall -o libplpgsql.so.1.0 \\\n\t\tpl_comp.o pl_exec.o pl_funcs.o pl_handler.o pl_parse.o -lc\n\tthe -bexpall makes it easier, but is new since AIX 4.2, so those with 3.2 or 4.1 still \n\tneed to do the .exp stuff\n\ncheers :-)\nAndreas\n\n",
"msg_date": "Wed, 28 Oct 1998 10:54:04 +0100",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "AIX 4.2.1 regression test"
},
{
"msg_contents": "> AIX 4.2.1 regression tests are ok :-)\n\nGreat. Will mark ya' down...\n\n> inet .. failed expected/inet.out is incorrect (ok, but please remake)\n\nD'Arcy/Paul/inet gang: Could you look at this and see if the inet.out is\nincorrect for your platforms too? I just took the results/inet.out,\ninspected it for gross errors (without knowing what I'm looking for) and\nput it in as the expected result. Don't know what might be wrong with\nit.\n\n> rules .. failed other sort order in select * from rtest_admin (guess ok)\n\nI think HPUX sees this too; it depends on the implementation of qsort()\non your platform.\n\n> modify configure test for cpp stdin\n> currently does xlc -E and fails to notice, that it does not work\n\nI'm not sure how easy it will be, or if we need to be frozen for this\nrelease. Should be a patch if it doesn't make it into the main v6.4.\nYou've already made a suggestion on what to look for, right? I'm\nrecalling that you had some discussion on the list about that already.\n\n - Tom\n",
"msg_date": "Wed, 28 Oct 1998 15:12:36 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] AIX 4.2.1 regression test"
},
{
"msg_contents": "Andreas Zeugswetter <[email protected]> writes:\n> AIX 4.2.1 regression tests are ok :-)\n> rules .. failed\t\tother sort order in select * from rtest_admin (guess ok)\n\nAh-hah, so HPUX is not the only platform where qsort chooses to output\nthose tuples in the other order. I feel better now ;-)\n\nI will go ahead and modify the rules test to do an \"order by\" in\nthe select * from rtest_admin command, so that it generates\npredictable results.\n\n> modify configure test for cpp stdin\n> \tcurrently does xlc -E and fails to notice, that it does not work\n\nI think we have two choices here:\n\n1. Try to improve configure's test for cpp-from-stdin some more, and add\nto it the idea that it might have to fall back to calling /lib/cpp\ndirectly if neither \"$(CPP)\" nor \"$(CPP) -\" work for reading from stdin.\n\n2. Rewrite the shell scripts that are using this feature so that they\ndon't need cpp-from-stdin, but just use a temporary file and plain\n$(CPP). Then we can forget about testing for it in configure.\n\nA quick \"glimpse\" shows there are only two shell scripts using\n$(CPPSTDIN), so I think choice #2 is the way to go. At this point in\nthe schedule, we need a high-probability-of-success fix, and tweaking an\nautoconf test is never a high-probability affair until you've actually\nrun it on a lot of platforms. Hacking the scripts might be ugly,\nbut it will work.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 10:50:35 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] AIX 4.2.1 regression test "
},
{
"msg_contents": "> > AIX 4.2.1 regression tests are ok :-)\n> > rules .. failed \n> Ah-hah, so HPUX is not the only platform where qsort chooses to output\n> those tuples in the other order. I feel better now ;-)\n\nDon't feel too good. There is a reason AIX is pronounced \"aches\" :)\n\n> > modify configure test for cpp stdin\n> 2. Rewrite the shell scripts that are using this feature so that they\n> don't need cpp-from-stdin, but just use a temporary file and plain\n> $(CPP). Then we can forget about testing for it in configure.\n> A quick \"glimpse\" shows there are only two shell scripts using\n> $(CPPSTDIN), so I think choice #2 is the way to go.\n\nI agree. Those two scripts use temporary files for other stages of their\nprocessing, so it isn't a change from what they already do. I was\ntempted to do that a month ago when I first ran across the problem of\nthe hardcoded cpp references. Are you going to try to do it now, or\nshould we wait for v6.4.1? I would think that it has a high probability\nof success, with testing from just one or two people.\n",
"msg_date": "Wed, 28 Oct 1998 16:31:53 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] AIX 4.2.1 regression test"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> I agree. Those two scripts use temporary files for other stages of their\n> processing, so it isn't a change from what they already do. I was\n> tempted to do that a month ago when I first ran across the problem of\n> the hardcoded cpp references. Are you going to try to do it now, or\n> should we wait for v6.4.1? I would think that it has a high probability\n> of success, with testing from just one or two people.\n\nI was planning to fix it now, unless Marc threatens to break my thumbs.\n\nBuild failures are a sufficiently important problem that I think we\noughta fix them ... and while this has only been reported on AIX,\nwe don't know for sure that it won't happen anywhere else.\n\nBesides, I think Marc is gonna be too busy breaking Bruce's thumbs\nto worry about me ;-).\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 12:17:41 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] AIX 4.2.1 regression test "
},
{
"msg_contents": "Thus spake Thomas G. Lockhart\n> > inet .. failed expected/inet.out is incorrect (ok, but please remake)\n> \n> D'Arcy/Paul/inet gang: Could you look at this and see if the inet.out is\n> incorrect for your platforms too? I just took the results/inet.out,\n> inspected it for gross errors (without knowing what I'm looking for) and\n> put it in as the expected result. Don't know what might be wrong with\n> it.\n\nAs discussed in irc, there seems to be a missing patch. I resent it\nto Bruce directly and he applied it. Please rerun the queries and\nreplace the expected output.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Wed, 28 Oct 1998 23:47:37 -0500 (EST)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] AIX 4.2.1 regression test"
}
] |
[
{
"msg_contents": "Hi All,\n\nI'm trying to compile the 6.3.2 version con SCO Openserver 5.0.4 with gcc\nprovided\nby SCO. It compile cleanly after some patches but I'm unable to connect to\nthe template1\ndatabase with postgres user. Psql reports \"User authentication failed\" etc.\nSo I want know\nwhere can I get the last binaries for SCO OpenServer.\n\nPlease reply directly because currently I'm not subscribed to the list.\n\nThanks,\n\nRoberto Fichera.\n\n",
"msg_date": "Wed, 28 Oct 1998 13:31:03 +0100",
"msg_from": "\"TeknSOFT S.n.c.\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "SCO 5.0.x binaries ?"
}
] |
[
{
"msg_contents": "Hi all,\n\nI don't know if this was alredy reported, if so, please forget it.\nI created a table with a serial data type and PostgreSQL creates an\nimplicit sequence, \nthen I droped the table but the sequence is already there, if I try to\nre-create \nthe same table PostgreSQL refuse to do that...\n\nprova=> create table a (i int, s smallint, c serial);\nNOTICE: CREATE TABLE will create implicit sequence a_c_seq for SERIAL\ncolumn a.c\nNOTICE: CREATE TABLE/UNIQUE will create implicit index a_c_key for\ntable a\nCREATE\nprova=> drop table a;\nDROP\nprova=> create table a (i int, s smallint, c serial);\nNOTICE: CREATE TABLE will create implicit sequence a_c_seq for SERIAL\ncolumn a.c\nNOTICE: CREATE TABLE/UNIQUE will create implicit index a_c_key for\ntable a\nERROR: a_c_seq relation already exists\n\nJose'\n",
"msg_date": "Wed, 28 Oct 1998 14:27:27 +0100",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "serial type"
},
{
"msg_contents": "> I don't know if this was alredy reported, if so, please forget it.\n> I created a table with a serial data type and PostgreSQL creates an\n> implicit sequence,\n> then I droped the table but the sequence is already there, if I try to\n> re-create\n> the same table PostgreSQL refuse to do that...\n\nYes, it is a known feature for v6.4, and is in the docs (which doesn't\nmake it right, of course :)\n\nHopefully we will merge this with other improvements for v6.5 and get it\nbetter then.\n\n - Tom\n",
"msg_date": "Wed, 28 Oct 1998 16:33:36 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] serial type"
}
] |
[
{
"msg_contents": "Hi all,\n\nI would to report that LOCK TABLE doesn't work outside\ntransactions but PostgreSQL accepts this statement outside\ntransactions whithout saying nothing...\n\nprova=> begin work;\nBEGIN\nprova=> select * from films where code = 'BL102';\ncode |title |did| date_prod|kind |len\n-----+-----------------+---+----------+----------+-----\nBL102|The African Queen|101|1951-08-11|Drama |01:43\n(1 row)\n\nprova=> commit work;\nEND\nprova=> lock table films;\nDELETE 0\n\nJose'\n",
"msg_date": "Wed, 28 Oct 1998 14:49:21 +0100",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "lock table"
},
{
"msg_contents": "> Hi all,\n> \n> I would to report that LOCK TABLE doesn't work outside\n> transactions but PostgreSQL accepts this statement outside\n> transactions whithout saying nothing...\n> \n> prova=> begin work;\n> BEGIN\n> prova=> select * from films where code = 'BL102';\n> code |title |did| date_prod|kind |len\n> -----+-----------------+---+----------+----------+-----\n> BL102|The African Queen|101|1951-08-11|Drama |01:43\n> (1 row)\n> \n> prova=> commit work;\n> END\n> prova=> lock table films;\n> DELETE 0\n> \n> Jose'\n\nGood point. I should fix that someday.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 11:20:43 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] lock table"
},
{
"msg_contents": "Added to TODO.\n\n\n> Hi all,\n> \n> I would to report that LOCK TABLE doesn't work outside\n> transactions but PostgreSQL accepts this statement outside\n> transactions whithout saying nothing...\n> \n> prova=> begin work;\n> BEGIN\n> prova=> select * from films where code = 'BL102';\n> code |title |did| date_prod|kind |len\n> -----+-----------------+---+----------+----------+-----\n> BL102|The African Queen|101|1951-08-11|Drama |01:43\n> (1 row)\n> \n> prova=> commit work;\n> END\n> prova=> lock table films;\n> DELETE 0\n> \n> Jose'\n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Fri, 11 Dec 1998 11:51:44 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] lock table"
}
] |
[
{
"msg_contents": "\nConsider this one what is to be released in 3 days...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 28 Oct 1998 11:33:59 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "BETA3 available..."
}
] |
[
{
"msg_contents": "Here are the open items. Thanks to Jan, the only 'hot' item left is the\nps args issue. People on non-BSD platforms will see all their backends\ncalled 'postmaster', because argv[0] changes do not reflect in ps arg\ndisplays.\n\nI have asked that at least we get set_proctitle() working for Linux, but\nno one wants to do it, and we can't remove the it because we would have\nto add the exec() back into backend creation, which is impossible at\nthis point.\n\nAt this point, I am not even sure if Marc will allow a fix so late in\nthe game. Perhaps we can put it into a minor 6.4 release.\n\nAs long as you don't think we are going to get tons of complaints, I am\nnot worried about it.\n\nEverything else is minor, so we are ready for 6.4 on November 1.\n\n---------------------------------------------------------------------------\n\n\nAdditions\n---------\nregression test all platforms\n\nSerious Items\n------------\nchange pg args for platforms that don't support argv changes\n\t(setproctitle()?, sendmail hack?)\n\nDocs\n----\ngenerate html/postscript documentation\n(User's Guide, Administrator's Guide, Programmer's Guide) (Thomas)\nmake sure all changes are documented properly\nmarkup and merge JDBC docs from Peter (Thomas, others??)\nmerge the release notes into doc/src/sgml/release.sgml (Bruce, Thomas)\ngenerate new text-format INSTALL and README from sgml sources (Thomas)\nmarkup of Jan's PL docs\n\nMinor items\n-----------\ncnf-ify still can exhaust memory, make SET KSQO more generic\npermissions on indexes: what do they do? should it be prevented?\nallow multiple generic operators in expressions without the use of parentheses\ndocument/trigger/rule so changes to pg_shadow create pg_pwd\nlarge objects orphanage\nimprove group handling\nimprove PRIMARY KEY handling\ngenerate postmaster pid file and remove flock/fcntl lock code\nadd ability to specifiy location of lock/socket files\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 11:16:42 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Open 6.4 items"
},
{
"msg_contents": "> >\n> > Here are the open items. Thanks to Jan, the only 'hot' item left is the\n> > ps args issue. People on non-BSD platforms will see all their backends\n> > called 'postmaster', because argv[0] changes do not reflect in ps arg\n> > displays.\n> >\n> > I have asked that at least we get set_proctitle() working for Linux, but\n> > no one wants to do it, and we can't remove the it because we would have\n> > to add the exec() back into backend creation, which is impossible at\n> > this point.\n> \n> I'm running Linux 2.1.88 and get\n> \n> 15572 p2 S 0:01 /usr/local/pgsql/bin/postmaster -o -F\n> 16121 p2 S 0:01 /usr/local/pgsql/bin/postgres localhost twieck twieck idle\n> \n> from ps. So what isn't working?\n\nWhoh, this is a shock. I thought this trick did not work under Linux. \nIf it does, and no one has complained, we can consider the issue closed.\n\nMassimo did a nice job of moving my ps status stuff into pg_status.h,\nbut it looks like the same code that modifies argv[0]. Perhaps it is\nthe Linux version you are using.\n\nIt give us a cheap 'db monitor' capability, wrapped into the 'ps'\ncommand.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 11:49:59 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": ">\n> Here are the open items. Thanks to Jan, the only 'hot' item left is the\n> ps args issue. People on non-BSD platforms will see all their backends\n> called 'postmaster', because argv[0] changes do not reflect in ps arg\n> displays.\n>\n> I have asked that at least we get set_proctitle() working for Linux, but\n> no one wants to do it, and we can't remove the it because we would have\n> to add the exec() back into backend creation, which is impossible at\n> this point.\n\n I'm running Linux 2.1.88 and get\n\n 15572 p2 S 0:01 /usr/local/pgsql/bin/postmaster -o -F\n 16121 p2 S 0:01 /usr/local/pgsql/bin/postgres localhost twieck twieck idle\n\n from ps. So what isn't working?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Wed, 28 Oct 1998 17:53:03 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> >\n> > Here are the open items. Thanks to Jan, the only 'hot' item left is the\n> > ps args issue. People on non-BSD platforms will see all their backends\n> > called 'postmaster', because argv[0] changes do not reflect in ps arg\n> > displays.\n> >\n> \n> Bruce,\n> \n> I asked for it a while ago but forgot about it. Anyway - I\n> think it is better to have precreated gram.c, y.tab.h and\n> scan.c files in src/pl/plpgsql/src too. Otherwise ppl not\n> having bison/flex might have a build problem.\n> \n> The only thing required is to take them out of the 'clean' rm\n> in Makefile.in and add the bison/flex created files to CVS.\n> If gram.c, y.tab.h and scan.l are present and newer than\n> gram.y and scan.l the Makefile will already skip the steps to\n> create them.\n\nDo they fail for people who have standard BSD yacc? Too large? No one\nhas complained about it, but it may be true.\n\nDone. Removed from Makefile.in, and added via cvs.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 12:06:23 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": ">\n> Here are the open items. Thanks to Jan, the only 'hot' item left is the\n> ps args issue. People on non-BSD platforms will see all their backends\n> called 'postmaster', because argv[0] changes do not reflect in ps arg\n> displays.\n>\n\nBruce,\n\n I asked for it a while ago but forgot about it. Anyway - I\n think it is better to have precreated gram.c, y.tab.h and\n scan.c files in src/pl/plpgsql/src too. Otherwise ppl not\n having bison/flex might have a build problem.\n\n The only thing required is to take them out of the 'clean' rm\n in Makefile.in and add the bison/flex created files to CVS.\n If gram.c, y.tab.h and scan.l are present and newer than\n gram.y and scan.l the Makefile will already skip the steps to\n create them.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Wed, 28 Oct 1998 18:06:55 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": " > I'm running Linux 2.1.88 and get\n > \n > 15572 p2 S 0:01 /usr/local/pgsql/bin/postmaster -o -F\n > 16121 p2 S 0:01 /usr/local/pgsql/bin/postgres localhost twieck twieck idle\n > \n > from ps. So what isn't working?\n\n Whoh, this is a shock. I thought this trick did not work under Linux. \n If it does, and no one has complained, we can consider the issue closed.\n\nFor whatever it's worth, I see the same status stuff changing under\nNetBSD 1.3.2. Seems to work fine.\n\nCheers,\nBrook\n",
"msg_date": "Wed, 28 Oct 1998 10:08:56 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> The only thing required is to take them out of the 'clean' rm\n> in Makefile.in and add the bison/flex created files to CVS.\n> If gram.c, y.tab.h and scan.l are present and newer than\n> gram.y and scan.l the Makefile will already skip the steps to\n> create them.\n\nWe should do that for the ecpg pre-processor also.\n\n - Tom\n",
"msg_date": "Wed, 28 Oct 1998 17:20:33 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> Here are the open items. Thanks to Jan, the only 'hot' item left is the\n> ps args issue. People on non-BSD platforms will see all their backends\n> called 'postmaster', because argv[0] changes do not reflect in ps arg\n> displays.\n\n> I have asked that at least we get set_proctitle() working for Linux, but\n> no one wants to do it, and we can't remove the it because we would have\n> to add the exec() back into backend creation, which is impossible at\n> this point.\n\nI think I was one of the louder complainers about the lack of proctitle\ninformation, but even I don't think it's worth holding up the release\nfor --- and it's certainly not worth taking a strong risk of breakage\nat this stage. Let's put it on the to-do-for-6.4.1 list.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 12:20:39 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "[email protected] (Jan Wieck) writes:\n> I asked for it a while ago but forgot about it. Anyway - I\n> think it is better to have precreated gram.c, y.tab.h and\n> scan.c files in src/pl/plpgsql/src too. Otherwise ppl not\n> having bison/flex might have a build problem.\n\nThere are quite a few places that need lex/yacc capability; plpgsql\nis not creating any new build requirement that did not exist before.\n(ecpg and bootparse are two examples I can think of offhand.)\nWe ship the main gram.c file not to avoid requiring lex/yacc, but\nbecause it is too big for some older yaccs.\n\nIt might be nice to eliminate the need for lex/yacc capability,\nbut post-beta3 is NOT the time to be doing \"might be nice\" stuff.\nLeave it be for now.\n\n\t\t\tregards, tom lane\n\nPS: BTW, I was able to build all the Postgres yacc files on a rather\nancient HPUX yacc, once I added enough -N switches. According to my\nnotes,\n\tYACC:/usr/bin/yacc\n\tYFLAGS:-d -Np2000 -Ns3000 -Nm100000 -Nl2000 -Na30000 -Nc10000\nworked with a pretty recent fileset. So I'm not sure we even really\nneed to distribute the main parser's gram.c. We could have configure\nplug in these switches if it fails to find bison...\n",
"msg_date": "Wed, 28 Oct 1998 12:30:50 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "> > Bruce,\n> >\n> > I asked for it a while ago but forgot about it. Anyway - I\n> > think it is better to have precreated gram.c, y.tab.h and\n> > scan.c files in src/pl/plpgsql/src too. Otherwise ppl not\n> > having bison/flex might have a build problem.\n> >\n> > The only thing required is to take them out of the 'clean' rm\n> > in Makefile.in and add the bison/flex created files to CVS.\n> > If gram.c, y.tab.h and scan.l are present and newer than\n> > gram.y and scan.l the Makefile will already skip the steps to\n> > create them.\n>\n> Do they fail for people who have standard BSD yacc? Too large? No one\n> has complained about it, but it may be true.\n\n They shouldn't be too large. But they get modified with\n sed(1) since this is a second independend scanner/parser\n inside the backend (after loading). I'm not 100% sure if the\n code generated by ANY other lex/yacc accepts the\n substitutions or if the resulting code is really that\n independet as it should be.\n\n>\n> Done. Removed from Makefile.in, and added via cvs.\n\n Thanks.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Wed, 28 Oct 1998 18:35:01 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Wed, 28 Oct 1998, Bruce Momjian wrote:\n\n> Here are the open items. Thanks to Jan, the only 'hot' item left is the\n> ps args issue. People on non-BSD platforms will see all their backends\n> called 'postmaster', because argv[0] changes do not reflect in ps arg\n> displays.\n\n\tSince there were no problem reports for this, I assumed that it\nwas an 'asthetic change' more then anything. Once v6.4 is released, I'll\ndive into it with the Linux guys here at the University...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 28 Oct 1998 15:17:45 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Doesn't work under Solaris, but, then again, neither does sendamil's, so\nthat isn't a shock :)\n\n\n\nOn Wed, 28 Oct 1998, Brook Milligan wrote:\n\n> > I'm running Linux 2.1.88 and get\n> > \n> > 15572 p2 S 0:01 /usr/local/pgsql/bin/postmaster -o -F\n> > 16121 p2 S 0:01 /usr/local/pgsql/bin/postgres localhost twieck twieck idle\n> > \n> > from ps. So what isn't working?\n> \n> Whoh, this is a shock. I thought this trick did not work under Linux. \n> If it does, and no one has complained, we can consider the issue closed.\n> \n> For whatever it's worth, I see the same status stuff changing under\n> NetBSD 1.3.2. Seems to work fine.\n> \n> Cheers,\n> Brook\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 28 Oct 1998 15:22:51 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> [email protected] (Jan Wieck) writes:\n> > I asked for it a while ago but forgot about it. Anyway - I\n> > think it is better to have precreated gram.c, y.tab.h and\n> > scan.c files in src/pl/plpgsql/src too. Otherwise ppl not\n> > having bison/flex might have a build problem.\n> \n> There are quite a few places that need lex/yacc capability; plpgsql\n> is not creating any new build requirement that did not exist before.\n> (ecpg and bootparse are two examples I can think of offhand.)\n> We ship the main gram.c file not to avoid requiring lex/yacc, but\n> because it is too big for some older yaccs.\n> \n> It might be nice to eliminate the need for lex/yacc capability,\n> but post-beta3 is NOT the time to be doing \"might be nice\" stuff.\n> Leave it be for now.\n> \n> \t\t\tregards, tom lane\n> \n> PS: BTW, I was able to build all the Postgres yacc files on a rather\n> ancient HPUX yacc, once I added enough -N switches. According to my\n> notes,\n> \tYACC:/usr/bin/yacc\n> \tYFLAGS:-d -Np2000 -Ns3000 -Nm100000 -Nl2000 -Na30000 -Nc10000\n> worked with a pretty recent fileset. So I'm not sure we even really\n> need to distribute the main parser's gram.c. We could have configure\n> plug in these switches if it fails to find bison...\n\nAgain, too late to play with. BSD yacc doesn't have switches that\nenable larger tables, so you have to have gram.c. Does plpgsql have\ntables too large for BSD yacc? Can someone test it?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 21:35:26 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> On Wed, 28 Oct 1998, Bruce Momjian wrote:\n> \n> > Here are the open items. Thanks to Jan, the only 'hot' item left is the\n> > ps args issue. People on non-BSD platforms will see all their backends\n> > called 'postmaster', because argv[0] changes do not reflect in ps arg\n> > displays.\n> \n> \tSince there were no problem reports for this, I assumed that it\n> was an 'asthetic change' more then anything. Once v6.4 is released, I'll\n> dive into it with the Linux guys here at the University...\n\nLooks like it works for most people. I will remove it from the list,\nand see if someone complains.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 22:36:45 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> that isn't a shock :)\n\nMan, if sendmail's doesn't work, that is really broken.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 22:37:24 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Wed, 28 Oct 1998, Bruce Momjian wrote:\n\n> > Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> > that isn't a shock :)\n> \n> Man, if sendmail's doesn't work, that is really broken.\n\n\tSolaris just doesn't have any mechanisms to work around the\nlimitation, I guess *shrug* It really sucks when you want to SIGHUP the\n\"parent process\", which, under FreeBSD at least, is the one that states:\n-accepting connections, but under Solaris they are *all* the same :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 00:32:49 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Doesn't work under UnixWare, either.\n\nThe Hermit Hacker wrote:\n> Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> that isn't a shock :)\n> \n> \n> \n> On Wed, 28 Oct 1998, Brook Milligan wrote:\n> \n> > > I'm running Linux 2.1.88 and get\n> > > \n> > > 15572 p2 S 0:01 /usr/local/pgsql/bin/postmaster -o -F\n> > > 16121 p2 S 0:01 /usr/local/pgsql/bin/postgres localhost twieck twieck idle\n> > > \n> > > from ps. So what isn't working?\n> > \n> > Whoh, this is a shock. I thought this trick did not work under Linux. \n> > If it does, and no one has complained, we can consider the issue closed.\n> > \n> > For whatever it's worth, I see the same status stuff changing under\n> > NetBSD 1.3.2. Seems to work fine.\n> > \n> > Cheers,\n> > Brook\n> > \n> \n> Marc G. Fournier \n> Systems Administrator @ hub.org \n> primary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n> \n> \n\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Thu, 29 Oct 1998 01:41:12 -0500",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "Bruce Momjian wrote:\n> > Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> > that isn't a shock :)\n> \n> Man, if sendmail's doesn't work, that is really broken.\n\nIf you want strange, here is the output from the UnixWare 7 ps command with \nsome different options:\n\n$ ps -ef | fgrep post\n pgsql 10775 1 TS 80 0 01:34:12 ? 0:00 postmaster -S -i -o -F \n pgsql 10869 10775 TS 80 0 01:45:30 ? 0:00 postmaster -S -i -o -F \n\n$ ps -e | fgrep post\n 10775 TS 80 ? 0:00 postgres\n 10869 TS 80 ? 0:00 postgres\n\nNotice that the PIDs are the same even if the program name is different \ndepending on the use of the 'f' option.\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Thu, 29 Oct 1998 01:48:26 -0500",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "> Bruce Momjian wrote:\n> > > Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> > > that isn't a shock :)\n> > \n> > Man, if sendmail's doesn't work, that is really broken.\n> \n> If you want strange, here is the output from the UnixWare 7 ps command with \n> some different options:\n> \n> $ ps -ef | fgrep post\n> pgsql 10775 1 TS 80 0 01:34:12 ? 0:00 postmaster -S -i -o -F \n> pgsql 10869 10775 TS 80 0 01:45:30 ? 0:00 postmaster -S -i -o -F \n> \n> $ ps -e | fgrep post\n> 10775 TS 80 ? 0:00 postgres\n> 10869 TS 80 ? 0:00 postgres\n> \n> Notice that the PIDs are the same even if the program name is different \n> depending on the use of the 'f' option.\n\nWow, that really is strange. Seems like with the -f, it looks in one\nplace, as though it knows the child did not exec, so they params must be\nthe same, while with no -f, it looks in the proper area. The really\nweird thing is that even the postmaster is called postgres. That is\nbizarre.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 29 Oct 1998 02:07:59 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> Bruce Momjian wrote:\n> > > Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> > > that isn't a shock :)\n> > \n> > Man, if sendmail's doesn't work, that is really broken.\n> \n> If you want strange, here is the output from the UnixWare 7 ps command with \n> some different options:\n> \n> $ ps -ef | fgrep post\n> pgsql 10775 1 TS 80 0 01:34:12 ? 0:00 postmaster -S -i -o -F \n> pgsql 10869 10775 TS 80 0 01:45:30 ? 0:00 postmaster -S -i -o -F \n> \n> $ ps -e | fgrep post\n> 10775 TS 80 ? 0:00 postgres\n> 10869 TS 80 ? 0:00 postgres\n> \n> Notice that the PIDs are the same even if the program name is different \n> depending on the use of the 'f' option.\n\nThat is really bizarre. The postmaster name is changed to postgres too!\n\n\nAh, I now know why things work on Linux. Massimo did more than just\nparameterize my argv code. In include/utils/ps_status.h, he put in\nLinux-specific code to handle the ps args. Seems under Linux, only\nargv[0] can be modified for ps display, for some reason. He erases the\nother args, and puts everything in argv[0].\n\nUnder non-linux, he uses the argv[0-4] for various displays, like my\norignal code. The major issue is that for Linux, he writes directly\ninto argv[0] memory. He sets the normal i/o static parameters in the\nstart of the string, then marks the next position, and writes status\ninformation into there.\n\nIn non-Linux, we don't write into argv[0], but change argv[0..4] to\npoint to our own alloc'ed strings, where the memory sits in our own\naddress space, not in the proc 'environment' address space. Getting\n'ps' to display process address space strings is a real trick, so it is\nnot surprising some O/S's don't support it.\n\nHe may have done it all in argv[0] because he is expanding the size of\nargv[0], and if they are layed out sequentially in environment memory,\nthe changes would over-write each other. What I don't understand is why\nps when it goes to look at argv[1], does not see garbage because it may\nbe pointing into the middle of the argv[0] string.) Can a Linux user\nuse ps to display all the params, like 'ps -ef' and see if it is showing\nargv[1] as part of the argv[0] string? You may have to get a wide\ndisplay to see it all.\n\nThe great news is that it seems it is working to write directly into the\nenvironment address space under Linux, so we don't need set_proctitle,\nand it is very fast.\n\nNon-linux platforms are not writing directly to environment memory, but\nare using the argv[0..4] trick.\n\nFYI, I have re-exec'ed the postmaster in postmaster.h so it has 4 args.\nargc is passed as a value to the program, so you can't change it, and if\nyou don't do the re-exec, ps only looks at the argv[0], and under\nnon-linux, that is bad because we use argv[0...4].\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 29 Oct 1998 02:31:20 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, The Hermit Hacker wrote:\n> \tSolaris just doesn't have any mechanisms to work around the\n> limitation, I guess *shrug* It really sucks when you want to SIGHUP\n> the \"parent process\", which, under FreeBSD at least, is the one that\n> states: -accepting connections, but under Solaris they are *all* the\n> same :)\n\n$ ps -eaf\n UID PID PPID C STIME TTY TIME CMD\n root 0 0 0 Oct 12 ? 0:01 sched\n root 1 0 0 Oct 12 ? 0:15 /etc/init -\n...\n\nYou'll note the 'PPID' field.\n\n3 guesses what that stands for.\n\n-- \n| Matthew N. Dodd | 78 280Z | 75 164E | 84 245DL | FreeBSD/NetBSD/Sprite/VMS |\n| [email protected] | This Space For Rent | ix86,sparc,m68k,pmax,vax |\n| http://www.jurai.net/~winter | Are you k-rad elite enough for my webpage? |\n\n",
"msg_date": "Thu, 29 Oct 1998 10:24:36 -0500 (EST)",
"msg_from": "\"Matthew N. Dodd\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": ">\n> On Thu, 29 Oct 1998, The Hermit Hacker wrote:\n> > Solaris just doesn't have any mechanisms to work around the\n> > limitation, I guess *shrug* It really sucks when you want to SIGHUP\n> > the \"parent process\", which, under FreeBSD at least, is the one that\n> > states: -accepting connections, but under Solaris they are *all* the\n> > same :)\n>\n> $ ps -eaf\n> UID PID PPID C STIME TTY TIME CMD\n> root 0 0 0 Oct 12 ? 0:01 sched\n> root 1 0 0 Oct 12 ? 0:15 /etc/init -\n> ...\n>\n> You'll note the 'PPID' field.\n>\n> 3 guesses what that stands for.\n>\n\n Don't see how this is related to the topic - sorry.\n\n PPID is the parent process ID. sched has no parent (it's a\n kernel pseudo process) and init has sched as father. For all\n other processes the PPID is set to init's PID at the time\n their father dies (you'll see lot's of PPID=1).\n\n But this all has nothing to do with changing the CMD column\n of the ps output from inside a running process.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Thu, 29 Oct 1998 17:31:19 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, Jan Wieck wrote:\n> > On Thu, 29 Oct 1998, The Hermit Hacker wrote:\n> > > Solaris just doesn't have any mechanisms to work around the\n> > > limitation, I guess *shrug* It really sucks when you want to SIGHUP\n> > > the \"parent process\", which, under FreeBSD at least, is the one that\n> > > states: -accepting connections, but under Solaris they are *all* the\n> > > same :)\n> >\n> > $ ps -eaf\n> > UID PID PPID C STIME TTY TIME CMD\n> > root 0 0 0 Oct 12 ? 0:01 sched\n> > root 1 0 0 Oct 12 ? 0:15 /etc/init -\n> > ...\n> >\n> > You'll note the 'PPID' field.\n> >\n> > 3 guesses what that stands for.\n> >\n> \n> Don't see how this is related to the topic - sorry.\n\nI really have to start explaining things using more words don't I.\n\n> PPID is the parent process ID. sched has no parent (it's a\n> kernel pseudo process) and init has sched as father. For all\n> other processes the PPID is set to init's PID at the time\n> their father dies (you'll see lot's of PPID=1).\n\nReread the bit of text I quoted. Read my reply. How does my reply\naddress the problem scrappy had?\n\nThe bits of ps output I quoted were only serving to demonstrate actual\ndata produced by the ps command I used. The actual commands weren't\nimportant, only the PID and PPID fields were.\n\n> But this all has nothing to do with changing the CMD column\n> of the ps output from inside a running process.\n\nNo, but changing the CMD column is only eye-candy. Its not necessary to\nbe able to tell which processes are children and which is the parent.\n\n-- \n| Matthew N. Dodd | 78 280Z | 75 164E | 84 245DL | FreeBSD/NetBSD/Sprite/VMS |\n| [email protected] | This Space For Rent | ix86,sparc,m68k,pmax,vax |\n| http://www.jurai.net/~winter | Are you k-rad elite enough for my webpage? |\n\n",
"msg_date": "Thu, 29 Oct 1998 11:43:59 -0500 (EST)",
"msg_from": "\"Matthew N. Dodd\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> Wow, that really is strange. Seems like with the -f, it looks in one\n> place, as though it knows the child did not exec, so they params must be\n> the same, while with no -f, it looks in the proper area. The really\n> weird thing is that even the postmaster is called postgres. That is\n> bizarre.\n\nYes. With -f, it looks in the process's argv... without, it looks in the\ntask struct where only the executable name was stored (no path or anything\nelse). It's a feature to prevent processes that spoof as something else...\n(not very effective, IMHO)\n\nTaral\n\n",
"msg_date": "Thu, 29 Oct 1998 11:05:38 -0600",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, Billy G. Allie wrote:\n\n> Doesn't work under UnixWare, either.\n> \n\n\tHow about sendmail? That is \"broken\" too?\n\n\n> The Hermit Hacker wrote:\n> > Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> > that isn't a shock :)\n> > \n> > \n> > \n> > On Wed, 28 Oct 1998, Brook Milligan wrote:\n> > \n> > > > I'm running Linux 2.1.88 and get\n> > > > \n> > > > 15572 p2 S 0:01 /usr/local/pgsql/bin/postmaster -o -F\n> > > > 16121 p2 S 0:01 /usr/local/pgsql/bin/postgres localhost twieck twieck idle\n> > > > \n> > > > from ps. So what isn't working?\n> > > \n> > > Whoh, this is a shock. I thought this trick did not work under Linux. \n> > > If it does, and no one has complained, we can consider the issue closed.\n> > > \n> > > For whatever it's worth, I see the same status stuff changing under\n> > > NetBSD 1.3.2. Seems to work fine.\n> > > \n> > > Cheers,\n> > > Brook\n> > > \n> > \n> > Marc G. Fournier \n> > Systems Administrator @ hub.org \n> > primary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n> > \n> > \n> \n> -- \n> ____ | Billy G. Allie | Domain....: [email protected]\n> | /| | 7436 Hartwell | Compuserve: 76337,2061\n> |-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n> |/ |LLIE | (313) 582-1540 | \n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 13:19:50 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "On Thu, 29 Oct 1998, Matthew N. Dodd wrote:\n\n> On Thu, 29 Oct 1998, The Hermit Hacker wrote:\n> > \tSolaris just doesn't have any mechanisms to work around the\n> > limitation, I guess *shrug* It really sucks when you want to SIGHUP\n> > the \"parent process\", which, under FreeBSD at least, is the one that\n> > states: -accepting connections, but under Solaris they are *all* the\n> > same :)\n> \n> $ ps -eaf\n> UID PID PPID C STIME TTY TIME CMD\n> root 0 0 0 Oct 12 ? 0:01 sched\n> root 1 0 0 Oct 12 ? 0:15 /etc/init -\n> ...\n> \n> You'll note the 'PPID' field.\n> \n> 3 guesses what that stands for.\n\n\tOkay, now you risk getting on my bad side :) I know what PPID\nstands for...now you tell me which of these processes to SIGHUP:\n\n root 18942 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n root 18946 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n root 18948 1 0 13:22:04 ? 0:00 /usr/local/sbin/sendmail\n root 22213 1 0 Oct 26 ? 1:40 /usr/local/sbin/sendmail\n\n\tAnd ya, I know, the one with the older date...the point is that\nyou can't really automate this, except to do:\n\n\tkill -HUP `ps -aef | grep sendmail | awk '{print $2}'`\n\n\tAnd SIGHUP them all...\n\n\n",
"msg_date": "Thu, 29 Oct 1998 13:24:06 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, Jan Wieck wrote:\n\n> >\n> > On Thu, 29 Oct 1998, The Hermit Hacker wrote:\n> > > Solaris just doesn't have any mechanisms to work around the\n> > > limitation, I guess *shrug* It really sucks when you want to SIGHUP\n> > > the \"parent process\", which, under FreeBSD at least, is the one that\n> > > states: -accepting connections, but under Solaris they are *all* the\n> > > same :)\n> >\n> > $ ps -eaf\n> > UID PID PPID C STIME TTY TIME CMD\n> > root 0 0 0 Oct 12 ? 0:01 sched\n> > root 1 0 0 Oct 12 ? 0:15 /etc/init -\n> > ...\n> >\n> > You'll note the 'PPID' field.\n> >\n> > 3 guesses what that stands for.\n> >\n> \n> Don't see how this is related to the topic - sorry.\n> \n> PPID is the parent process ID. sched has no parent (it's a\n> kernel pseudo process) and init has sched as father. For all\n> other processes the PPID is set to init's PID at the time\n> their father dies (you'll see lot's of PPID=1).\n> \n> But this all has nothing to do with changing the CMD column\n> of the ps output from inside a running process.\n\nIn Matthew's defence, I think the point he was trying to bring across was\nthat you should be able to look at hte PPID. sendmail, when you start,\ntends to list its PPID as '1'...but, as I showed in my last email, that\ndoesn't appear to be \"unique\"...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 13:26:23 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> Bruce Momjian wrote:\n> > > Doesn't work under Solaris, but, then again, neither does sendamil's, so\n> > > that isn't a shock :)\n> > \n> > Man, if sendmail's doesn't work, that is really broken.\n> \n> If you want strange, here is the output from the UnixWare 7 ps command with \n> some different options:\n> \n> $ ps -ef | fgrep post\n> pgsql 10775 1 TS 80 0 01:34:12 ? 0:00 postmaster -S -i -o -F \n> pgsql 10869 10775 TS 80 0 01:45:30 ? 0:00 postmaster -S -i -o -F \n> \n> $ ps -e | fgrep post\n> 10775 TS 80 ? 0:00 postgres\n> 10869 TS 80 ? 0:00 postgres\n\nMy guess is that ps without -f looks at the inode that opened the file,\nand postmaster and postgres have the same inode because postmaster is\nsymlinked to postgres.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 29 Oct 1998 12:34:59 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> root 18942 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> root 18946 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> root 18948 1 0 13:22:04 ? 0:00 /usr/local/sbin/sendmail\n> root 22213 1 0 Oct 26 ? 1:40 /usr/local/sbin/sendmail\n\nEr... if you have this, you have a problem. This means there are TWO\nsendmail daemons running... There should only be ONE with a ppid of 1. It\nlooks like 18948 is blah.\n\nTaral\n\n",
"msg_date": "Thu, 29 Oct 1998 11:36:02 -0600",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, The Hermit Hacker wrote:\n> \tOkay, now you risk getting on my bad side :) I know what PPID\n> stands for...now you tell me which of these processes to SIGHUP:\n> \n> root 18942 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> root 18946 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n\n2 open connections.\n\n> root 18948 1 0 13:22:04 ? 0:00 /usr/local/sbin/sendmail\n\na local delivery spawned by an open connection that closed and left the\nlocal delivery to finish, thus orphaning it to init (PID1)\n\n> root 22213 1 0 Oct 26 ? 1:40 /usr/local/sbin/sendmail\n\nThe listener.\n\n> \tAnd ya, I know, the one with the older date...the point is that\n> you can't really automate this, except to do:\n\nYou're trying to kill the listener?\n\nYou do know you can use truss and lsof to figure this out as well right?\n\nOr you could do:\n\n\tkill -HUP `head -1 /var/run/sendmail.pid`\n\nAlso, check out what the BSD style ps shows under Solaris (/usr/ucb/ps).\n\n-- \n| Matthew N. Dodd | 78 280Z | 75 164E | 84 245DL | FreeBSD/NetBSD/Sprite/VMS |\n| [email protected] | This Space For Rent | ix86,sparc,m68k,pmax,vax |\n| http://www.jurai.net/~winter | Are you k-rad elite enough for my webpage? |\n\n",
"msg_date": "Thu, 29 Oct 1998 12:40:26 -0500 (EST)",
"msg_from": "\"Matthew N. Dodd\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, Taral wrote:\n> > root 18942 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> > root 18946 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> > root 18948 1 0 13:22:04 ? 0:00 /usr/local/sbin/sendmail\n> > root 22213 1 0 Oct 26 ? 1:40 /usr/local/sbin/sendmail\n> \n> Er... if you have this, you have a problem. This means there are TWO\n> sendmail daemons running... There should only be ONE with a ppid of 1. It\n> looks like 18948 is blah.\n\nNo, it looks like there are 4 sendmail daemons running.\n\nThe issue is to determine what each of them are doing.\n\ndaemon != listener\n\n-- \n| Matthew N. Dodd | 78 280Z | 75 164E | 84 245DL | FreeBSD/NetBSD/Sprite/VMS |\n| [email protected] | This Space For Rent | ix86,sparc,m68k,pmax,vax |\n| http://www.jurai.net/~winter | Are you k-rad elite enough for my webpage? |\n\n",
"msg_date": "Thu, 29 Oct 1998 12:43:18 -0500 (EST)",
"msg_from": "\"Matthew N. Dodd\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> I really have to start explaining things using more words don't I.\n\nOnly if you want anyone to bother paying attention to you.\n\n - Thomas (who didn't)\n",
"msg_date": "Thu, 29 Oct 1998 18:17:59 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, Taral wrote:\n\n> > root 18942 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> > root 18946 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> > root 18948 1 0 13:22:04 ? 0:00 /usr/local/sbin/sendmail\n> > root 22213 1 0 Oct 26 ? 1:40 /usr/local/sbin/sendmail\n> \n> Er... if you have this, you have a problem. This means there are TWO\n> sendmail daemons running... There should only be ONE with a ppid of 1. It\n> looks like 18948 is blah.\n\nGood guess, but that is normal behaviour, at least under Solaris :) It\ntook me several 'ps -aef's to actually find it duplicated, it isn't always\nthere.\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 22:02:42 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Thu, 29 Oct 1998, Matthew N. Dodd wrote:\n\n> On Thu, 29 Oct 1998, The Hermit Hacker wrote:\n> > \tOkay, now you risk getting on my bad side :) I know what PPID\n> > stands for...now you tell me which of these processes to SIGHUP:\n> > \n> > root 18942 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> > root 18946 22213 0 13:22:03 ? 0:00 /usr/local/sbin/sendmail\n> \n> 2 open connections.\n> \n> > root 18948 1 0 13:22:04 ? 0:00 /usr/local/sbin/sendmail\n> \n> a local delivery spawned by an open connection that closed and left the\n> local delivery to finish, thus orphaning it to init (PID1)\n> \n> > root 22213 1 0 Oct 26 ? 1:40 /usr/local/sbin/sendmail\n> \n> The listener.\n> \n> > \tAnd ya, I know, the one with the older date...the point is that\n> > you can't really automate this, except to do:\n> \n> You're trying to kill the listener?\n> \n> You do know you can use truss and lsof to figure this out as well right?\n> \n> Or you could do:\n> \n> \tkill -HUP `head -1 /var/run/sendmail.pid`\n\n\tYes, of course, but this was not the point of the whole\nthread...:)\n\n> Also, check out what the BSD style ps shows under Solaris (/usr/ucb/ps).\n\n\tNo apparent difference in the output...wait, you'll take me too\nliterally on that, eh? :) Both show the sasme information in the last\nfield../usr/local/sbin/sendmail -bd -q1h :)\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 22:06:25 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Hallo Jan,\n\nDo I remember right that your pathes to speed up ORDER BYs (by \nomitting them when not needed) did not make it into 6.4 .\n\nIf that is the case, are they available anywhere ?\n\nI really need them (fast) for my new project.\n\n-------------\nHannu Krosing\n",
"msg_date": "Fri, 30 Oct 1998 18:38:55 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": false,
"msg_subject": "ORDER BY optimisations"
},
{
"msg_contents": "> Hallo Jan,\n> \n> Do I remember right that your pathes to speed up ORDER BYs (by \n> omitting them when not needed) did not make it into 6.4 .\n> \n> If that is the case, are they available anywhere ?\n> \n> I really need them (fast) for my new project.\n\nLIMIT will probably be added to 6.4.1. Queries that use '%text%' can\nnot use indexes because they are not anchored at the beginning. \nfulltextindex is in contrib for those cases.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Fri, 30 Oct 1998 12:19:40 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] ORDER BY optimisations"
},
{
"msg_contents": "Bruce Momjian wrote:\n> \n> > Hallo Jan,\n> >\n> > Do I remember right that your pathes to speed up ORDER BYs (by\n> > omitting them when not needed) did not make it into 6.4 .\n> >\n> > If that is the case, are they available anywhere ?\n> >\n> > I really need them (fast) for my new project.\n> \n> LIMIT will probably be added to 6.4.1. \n\n\nActually I don't need LIMIT that much (for me using CURSOR/MOVE/FETCH \nis quite ok). The main benefit from LIMIT seems to be the ability \nof giving the (future) optimiser a hint that we actulally need only \na small part of the whole query so it may be better to use an index.\n\nWhat I am after, is the patch that removed the redundant sort node \nwhen the access is already by an index matching the sort.\n\n> Queries that use '%text%' can not use indexes because they are not \n> anchored at the beginning.\n> fulltextindex is in contrib for those cases.\n\nIt still seems a bit of a cludge, although a useful one as its usage \nis quite different from the use of other indexes. It also seems to be \nquite liberal with wasting space as it makes both an additional table\n_and_ an index for the words it indexes.\n\nIn fact I'm currently using my own fulltext indexing scheme outside \nthe database. I'm planning to work on including it in the pgsql \nbackend, once I figure out how the extending of access methods works.\n\n--------------\nHannu\n",
"msg_date": "Sat, 31 Oct 1998 16:53:59 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] ORDER BY optimisations"
},
{
"msg_contents": ">\n> Hallo Jan,\n>\n> Do I remember right that your pathes to speed up ORDER BYs (by\n> omitting them when not needed) did not make it into 6.4 .\n>\n> If that is the case, are they available anywhere ?\n>\n> I really need them (fast) for my new project.\n>\n> -------------\n> Hannu Krosing\n>\n>\n\n Yepp, it didn't made it.\n\n There where two different ones out, my and one from - hmmm -\n was it Tatsuo or Hinoue? My one only suppresses the final\n sort if\n\n 1. the plan is a Sort->IndexScan,\n\n 2. there is an ORDER BY clause,\n\n 3. the index choosen by the planner matches ALL attributes\n given in the ORDER BY clause (extra indexed attributes\n not in ORDER BY ignored),\n\n 4. and finally all sort operators are ASCENDING.\n\n There are many debugging printf()'s in the patch and I think\n one of them is still active while the others are commented\n out. You need to comment out the last one yourself after you\n found out that your queries are what causes it to suppress\n the sorts.\n\n Anyway, you said you need it fast, so here it is.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\ndiff -cr src.orig/backend/optimizer/plan/planner.c src/backend/optimizer/plan/planner.c\n*** src.orig/backend/optimizer/plan/planner.c\tWed Oct 14 19:12:36 1998\n--- src/backend/optimizer/plan/planner.c\tWed Oct 14 23:17:08 1998\n***************\n*** 48,53 ****\n--- 48,59 ----\n \n #include \"executor/executor.h\"\n \n+ #include \"utils/builtins.h\"\n+ #include \"utils/syscache.h\"\n+ #include \"access/genam.h\"\n+ #include \"parser/parse_oper.h\"\n+ \n+ static bool need_sortplan(List *sortcls, Plan *plan);\n static Plan *make_sortplan(List *tlist, List *sortcls, Plan *plannode);\n extern Plan *make_groupPlan(List **tlist, bool tuplePerGroup,\n \t\t\t List *groupClause, Plan *subplan);\n***************\n*** 281,292 ****\n \t}\n \telse\n \t{\n! \t\tif (parse->sortClause)\n \t\t\treturn make_sortplan(tlist, parse->sortClause, result_plan);\n \t\telse\n \t\t\treturn (Plan *) result_plan;\n \t}\n \n }\n \n /*\n--- 287,450 ----\n \t}\n \telse\n \t{\n! \t\tif (parse->sortClause && need_sortplan(parse->sortClause, result_plan))\n \t\t\treturn make_sortplan(tlist, parse->sortClause, result_plan);\n \t\telse\n \t\t\treturn (Plan *) result_plan;\n \t}\n \n+ }\n+ \n+ static TargetEntry *\n+ get_matching_tle(Plan *plan, Resdom *resdom)\n+ {\n+ \tList\t\t*i;\n+ \tTargetEntry\t*tle;\n+ \n+ \tforeach (i, plan->targetlist) {\n+ \t\ttle = (TargetEntry *)lfirst(i);\n+ \t\tif (tle->resdom->resno == resdom->resno)\n+ \t\t\treturn tle;\n+ \t}\n+ \treturn NULL;\n+ }\n+ \n+ static bool\n+ need_sortplan(List *sortcls, Plan *plan)\n+ {\n+ \tRelation\tindexRel;\n+ \tIndexScan\t*indexScan;\n+ \tOid\t\tindexId;\n+ \tList\t\t*i;\n+ \tHeapTuple\thtup;\n+ \tForm_pg_index\tindex_tup;\n+ \tint\t\tkey_no = 0;\n+ \n+ \t/*\n+ \tprintf(\"check if need_sortplan ... \");\n+ \t*/\n+ \n+ \tif (nodeTag(plan) != T_IndexScan) {\n+ \t\t/*\n+ \t\tprintf(\"not an index scan\\n\");\n+ \t\t*/\n+ \t\treturn TRUE;\n+ \t}\n+ \n+ \tindexScan = (IndexScan *)plan;\n+ \n+ \tif (plan->lefttree != NULL) {\n+ \t\t/*\n+ \t\tprintf(\"scan has lefttree\\n\");\n+ \t\t*/\n+ \t\treturn TRUE;\n+ \t}\n+ \tif (plan->righttree != NULL) {\n+ \t\t/*\n+ \t\tprintf(\"scan has righttree\\n\");\n+ \t\t*/\n+ \t\treturn TRUE;\n+ \t}\n+ \n+ \tif (length(indexScan->indxid) != 1) {\n+ \t\t/*\n+ \t\tprintf(\"scanning multiple indices\\n\");\n+ \t\t*/\n+ \t\treturn TRUE;\n+ \t}\n+ \n+ \tif (length(sortcls) > 8) {\n+ \t\t/*\n+ \t\tprintf(\"sort clause too long (>8)\\n\");\n+ \t\t*/\n+ \t\treturn TRUE;\n+ \t}\n+ \n+ \tindexId = lfirsti(indexScan->indxid);\n+ \n+ \tindexRel = index_open(indexId);\n+ \tif (strcmp(nameout(&(indexRel->rd_am->amname)), \"btree\") != 0) {\n+ \t\t/*\n+ \t\tprintf(\"not a btree index\\n\");\n+ \t\t*/\n+ \t\theap_close(indexRel);\n+ \t\treturn TRUE;\n+ \t}\n+ \theap_close(indexRel);\n+ \n+ \thtup = SearchSysCacheTuple(INDEXRELID,\n+ \t\t\tObjectIdGetDatum(indexId), 0, 0, 0);\n+ \tif (!HeapTupleIsValid(htup)) {\n+ \t\telog(ERROR, \"cache lookup for index %d failed\", indexId);\n+ \t}\n+ \tindex_tup = (Form_pg_index) GETSTRUCT(htup);\n+ \n+ \tforeach (i, sortcls) {\n+ \t\tSortClause\t*sortcl;\n+ \t\tResdom\t\t*resdom;\n+ \t\tTargetEntry\t*tle;\n+ \t\tVar\t\t*var;\n+ \n+ \t\tsortcl = (SortClause *) lfirst(i);\n+ \n+ \t\t/*\n+ \t\tprintf(\"\\nchecking sortclause %s\\n\", nodeToString(sortcl));\n+ \t\t*/\n+ \n+ \t\tresdom = sortcl->resdom;\n+ \t\ttle = get_matching_tle(plan, resdom);\n+ \t\tif (tle == NULL) {\n+ \t\t\t/*\n+ \t\t\tprintf(\"matching target entry not found\\n\");\n+ \t\t\t*/\n+ \t\t\treturn TRUE;\n+ \t\t}\n+ \t\tif (nodeTag(tle->expr) != T_Var) {\n+ \t\t\t/*\n+ \t\t\tprintf(\"target entry not a Var\\n\");\n+ \t\t\t*/\n+ \t\t\treturn TRUE;\n+ \t\t}\n+ \t\tvar = (Var *)(tle->expr);\n+ \n+ \t\tif (var->varno != indexScan->scan.scanrelid) {\n+ \t\t\t/*\n+ \t\t\tprintf(\"Var not from scanrelid\\n\");\n+ \t\t\t*/\n+ \t\t\treturn TRUE;\n+ \t\t}\n+ \n+ \t\tif (var->varattno != index_tup->indkey[key_no]) {\n+ \t\t\t/*\n+ \t\t\tprintf(\"attribute sorted does not match indexed att\\n\");\n+ \t\t\t*/\n+ \t\t\treturn TRUE;\n+ \t\t}\n+ \n+ \t\tif (oprid(oper(\"<\", resdom->restype, resdom->restype, FALSE)) != sortcl->opoid) {\n+ \t\t\t/*\n+ \t\t\tprintf(\"opoid should be %d - is %d\\n\",\n+ \t\t\t\toprid(oper(\"<\", resdom->restype, resdom->restype, FALSE)), sortcl->opoid);\n+ \t\t\t*/\n+ \t\t\treturn TRUE;\n+ \t\t}\n+ \n+ \t\tkey_no++;\n+ \t}\n+ \tif (key_no < 8 && index_tup->indkey[key_no] != 0) {\n+ \t\t/*\n+ \t\tprintf(\"there are more indexed fields! \");\n+ \t\t*/\n+ \t\treturn TRUE;\n+ \t}\n+ \n+ \tprintf(\"SUPPRESSING sort over index scan\\n\");\n+ \n+ \t/*\n+ \tprintf(\"scan = %s\\n\\n\", nodeToString(indexScan));\n+ \t*/\n+ \n+ \treturn FALSE;\n }\n \n /*\n",
"msg_date": "Sat, 31 Oct 1998 17:33:52 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] ORDER BY optimisations"
},
{
"msg_contents": "> 1. the plan is a Sort->IndexScan,\n>\n> 2. there is an ORDER BY clause,\n>\n> 3. the index choosen by the planner matches ALL attributes\n> given in the ORDER BY clause (extra indexed attributes\n> not in ORDER BY ignored),\n\n Ooops - sorry. Took another look at the patch and saw, that\n it actually does not ignore extra attributes in the index.\n Maybe you want to force sort suppression then too and comment\n out the 'return TRUE' for this case.\n\n BTW: This or an enhanced version (suppressing more stupid\n sort cases) is my first candidate for the v6.4.1 feature\n patch.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Sat, 31 Oct 1998 18:07:04 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] ORDER BY optimisations"
}
] |
[
{
"msg_contents": "\n> I'm running Linux 2.1.88 and get\n>\n> 15572 p2 S 0:01 /usr/local/pgsql/bin/postmaster -o -F\n> 16121 p2 S 0:01 /usr/local/pgsql/bin/postgres localhost twieck twieck idle\n>\n> from ps. So what isn't working?\n\nAlso works on AIX :-) \n\n\n",
"msg_date": "Wed, 28 Oct 1998 17:52:24 +0100",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] Open 6.4 items"
}
] |
[
{
"msg_contents": "I did a database dump and reload on my play system just now in order\nto check Bruce's latest system table change. Something very weird\nhas happened since the prior dump, which was Saturday: two regular\ntables and two sequences lost all their tuples! (The empty sequences\nmade pg_dump unhappy, so I had to delete them just to make a dump.)\n\nI don't believe I did anything to touch those particular tables since\nSaturday. Certainly I did not do a DELETE.\n\nDisappearing data in what is supposed to be almost-releasable code\nis distressing, to put it mildly. Any thoughts on what to look at?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 14:49:51 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "This is disturbing..."
}
] |
[
{
"msg_contents": "I on a SUnOS 5.6 box without root access trying to compile Postgres 6.3.2\nand getting the following:\n\n\n\n--------------------------------------------------------------------\nYou must use GNU make to use Postgres. It may be installed\non your system with the name 'gmake'.\n\nNOTE: If you are sure that you are using GNU make and you are\n still getting this message, you may simply need to run\n the configure program.\n--------------------------------------------------------------------\n\nMy make command line: make all > & make.log &\n\nThe configure looked like: ./configure --prefix=/my_path/pgsql --with-template=sparc_solaris-gcc\n\nWhat am I doing wrong?\n\nThanks\n\n",
"msg_date": "Wed, 28 Oct 1998 14:41:26 -0600 (CST)",
"msg_from": "Andy Lewis <[email protected]>",
"msg_from_op": true,
"msg_subject": "Compile Info"
}
] |
[
{
"msg_contents": "As long as we're doing last-minute changes that force initdb ...\n\nI believe I've just identified another error in pg_operator.h.\nSpecifically, this tuple is bogus and ought to be deleted:\n\nDATA(insert OID = 619 ( \"<\"\t PGUID 0 b t f 704 704 16 0 0 0 0 intervalct - - ));\n\nThis tuple defines \"<\" on tinterval's (type 704 = TINTERVAL) as\nbeing \"intervalct\", or interval-contained-in. Now that doesn't\nmake a lot of sense. Furthermore there is already a tuple\nthat defines \"<<\" as being intervalct:\n\nDATA(insert OID = 573 ( \"<<\"\t PGUID 0 b t f 704 704 16 0 0 0 0 intervalct - - ));\n\nand there is another tuple that defines \"<\" on tintervals as\nbeing \"intervallt\":\n\nDATA(insert OID = 813 ( \"<\"\t PGUID 0 b t f 704 704 16 813 0 0 0 intervallt - - ));\n\nI found this while trying to track down why the tinterval regression\ntest fails on HPUX (it produces the right tuples, but in the wrong\nsort order ... and unlike the rules case, the select does have an\nORDER BY clause). As far as I can tell from the CVS logs, the bogosity\nhas been in there for at least a year.\n\nIt looks to me like the bogus tuple is masking the correct definition\nof \"<\" and causing tintervals to sort incorrectly. Why this doesn't\nhappen on many other machines isn't clear, but maybe it has something\nto do with the platform-dependency of qsort that we already identified?\n\nAnyway, I'm a rank newbie at hacking this stuff, so I'm not about\nto commit this change without asking someone else to check my work.\nBut if it is indeed wrong, right now seems like a good time to fix it...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 15:49:55 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Another bug in pg_operator.h"
},
{
"msg_contents": "On Wed, 28 Oct 1998, Tom Lane wrote:\n\n> As long as we're doing last-minute changes that force initdb ...\n\nYou guys all know I'm going to be a super-prick next release, right?\n*grin* Just a pre-warning... :)\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 28 Oct 1998 17:07:32 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "Actually, the darn thing is *crawling* with bugs.\n\nOnce I understood what the commute and negate links were all about,\nI applied a few simple sanity checks (like, if A says it commutes to B,\ndoes B say it commutes to A?). This was, um, fruitful.\n\nThe comparison operators on booleans, tintervals, and lsegs are all\npretty seriously busted --- they have incorrect commute and/or negate\nlinks. I am worried about what other similar bugs may be in there,\nbut short of groveling over the entries by hand, I don't know what else\nto do to check the table further.\n\nBTW Bruce's fix of this morning wasn't right either --- he misspelled\nthe name of the lseg_ne comparison routine.\n\nI am currently running a build and regression test for the attached\ndiffs, which I propose to commit if the regress passes.\n\nI expect to find that this explains the tinterval regression differences\nI've been seeing. The tinterval expected file is clearly wrong,\nbecause it contains \"sorted\" lists of tintervals that are obviously not\nin order. I've been getting a different but just as wrong sort order\non HPUX. With luck this fix will produce a correct sort order.\n\nBTW, is it sufficient to edit pg_operator.h, rebuild, reinstall, and\ninitdb? Or is there some additional incantation that needs to be made\nwhen patching pg_operator.h?\n\n\t\t\tregards, tom lane\n\nPS: Stop the presses .... looks like int8 and int4-vs-int8 comparators are\nbogus too ... geez ...\n\n\n*** pg_operator.h.orig\tWed Oct 28 12:10:48 1998\n--- pg_operator.h\tWed Oct 28 18:32:20 1998\n***************\n*** 95,102 ****\n DATA(insert OID = 80 ( \"<=\"\t PGUID 0 b t f 23 20 16 419 419 0 0 int48le intlesel intlejoinsel ));\n DATA(insert OID = 82 ( \">=\"\t PGUID 0 b t f 23 20 16 418 418 0 0 int48ge intgesel intgejoinsel ));\n \n! DATA(insert OID = 58 ( \"<\"\t\t PGUID 0 b t f 16 16 16 85 91 0 0 boollt intltsel intltjoinsel ));\n! DATA(insert OID = 59 ( \">\"\t\t PGUID 0 b t f 16 16 16 85 91 0 0 boolgt intltsel intltjoinsel ));\n DATA(insert OID = 85 ( \"<>\"\t PGUID 0 b t f 16 16 16 85 91 0 0 boolne neqsel neqjoinsel ));\n DATA(insert OID = 91 ( \"=\"\t\t PGUID 0 b t t 16 16 16 91 85 0 0 booleq eqsel eqjoinsel ));\n #define BooleanEqualOperator 91\n--- 95,102 ----\n DATA(insert OID = 80 ( \"<=\"\t PGUID 0 b t f 23 20 16 419 419 0 0 int48le intlesel intlejoinsel ));\n DATA(insert OID = 82 ( \">=\"\t PGUID 0 b t f 23 20 16 418 418 0 0 int48ge intgesel intgejoinsel ));\n \n! DATA(insert OID = 58 ( \"<\"\t\t PGUID 0 b t f 16 16 16 59 0 0 0 boollt intltsel intltjoinsel ));\n! DATA(insert OID = 59 ( \">\"\t\t PGUID 0 b t f 16 16 16 58 0 0 0 boolgt intltsel intltjoinsel ));\n DATA(insert OID = 85 ( \"<>\"\t PGUID 0 b t f 16 16 16 85 91 0 0 boolne neqsel neqjoinsel ));\n DATA(insert OID = 91 ( \"=\"\t\t PGUID 0 b t t 16 16 16 91 85 0 0 booleq eqsel eqjoinsel ));\n #define BooleanEqualOperator 91\n***************\n*** 191,198 ****\n DATA(insert OID = 517 ( \"<->\"\t PGUID 0 b t f 600 600 701 517 0 0 0 point_distance intltsel intltjoinsel ));\n DATA(insert OID = 518 ( \"<>\"\t PGUID 0 b t f 23 23 16 518 96 0 0 int4ne neqsel neqjoinsel ));\n DATA(insert OID = 519 ( \"<>\"\t PGUID 0 b t f 21 21 16 519 94 0 0 int2ne neqsel neqjoinsel ));\n! DATA(insert OID = 520 ( \">\"\t PGUID 0 b t f 21 21 16 95 0 0 0 int2gt intgtsel intgtjoinsel ));\n! DATA(insert OID = 521 ( \">\"\t PGUID 0 b t f 23 23 16 97 0 0 0 int4gt intgtsel intgtjoinsel ));\n DATA(insert OID = 522 ( \"<=\"\t PGUID 0 b t f 21 21 16 524 520 0 0 int2le intltsel intltjoinsel ));\n DATA(insert OID = 523 ( \"<=\"\t PGUID 0 b t f 23 23 16 525 521 0 0 int4le intltsel intltjoinsel ));\n DATA(insert OID = 524 ( \">=\"\t PGUID 0 b t f 21 21 16 522 95 0 0 int2ge intgtsel intgtjoinsel ));\n--- 191,198 ----\n DATA(insert OID = 517 ( \"<->\"\t PGUID 0 b t f 600 600 701 517 0 0 0 point_distance intltsel intltjoinsel ));\n DATA(insert OID = 518 ( \"<>\"\t PGUID 0 b t f 23 23 16 518 96 0 0 int4ne neqsel neqjoinsel ));\n DATA(insert OID = 519 ( \"<>\"\t PGUID 0 b t f 21 21 16 519 94 0 0 int2ne neqsel neqjoinsel ));\n! DATA(insert OID = 520 ( \">\"\t PGUID 0 b t f 21 21 16 95 522 0 0 int2gt intgtsel intgtjoinsel ));\n! DATA(insert OID = 521 ( \">\"\t PGUID 0 b t f 23 23 16 97 523 0 0 int4gt intgtsel intgtjoinsel ));\n DATA(insert OID = 522 ( \"<=\"\t PGUID 0 b t f 21 21 16 524 520 0 0 int2le intltsel intltjoinsel ));\n DATA(insert OID = 523 ( \"<=\"\t PGUID 0 b t f 23 23 16 525 521 0 0 int4le intltsel intltjoinsel ));\n DATA(insert OID = 524 ( \">=\"\t PGUID 0 b t f 21 21 16 522 95 0 0 int2ge intgtsel intgtjoinsel ));\n***************\n*** 298,305 ****\n DATA(insert OID = 617 ( \"<->\"\t PGUID 0 b t f 601 603 701 0 0 0 0 dist_sb - - ));\n DATA(insert OID = 618 ( \"<->\"\t PGUID 0 b t f 600 602 701 0 0 0 0 dist_ppath - - ));\n \n- DATA(insert OID = 619 ( \"<\"\t PGUID 0 b t f 704 704 16 0 0 0 0 intervalct - - ));\n- \n DATA(insert OID = 620 ( \"=\"\t PGUID 0 b t t 700 700\t16 620 621\t622 622 float4eq eqsel eqjoinsel ));\n DATA(insert OID = 621 ( \"<>\"\t PGUID 0 b t f 700 700\t16 621 620\t0 0 float4ne neqsel neqjoinsel ));\n DATA(insert OID = 622 ( \"<\"\t PGUID 0 b t f 700 700\t16 623 625\t0 0 float4lt intltsel intltjoinsel ));\n--- 298,303 ----\n***************\n*** 400,411 ****\n DATA(insert OID = 808 ( \"?-\"\t PGUID 0 b t f 600 600\t 16 808 0 0 0 point_horiz - - ));\n DATA(insert OID = 809 ( \"?|\"\t PGUID 0 b t f 600 600\t 16 809 0 0 0 point_vert - - ));\n \n! DATA(insert OID = 811 ( \"=\"\t PGUID 0 b t t 704 704 16 811 0 0 0 intervaleq - - ));\n! DATA(insert OID = 812 ( \"<>\"\t PGUID 0 b t f 704 704 16 812 0 0 0 intervalne - - ));\n! DATA(insert OID = 813 ( \"<\"\t PGUID 0 b t f 704 704 16 813 0 0 0 intervallt - - ));\n! DATA(insert OID = 814 ( \">\"\t PGUID 0 b t f 704 704 16 814 0 0 0 intervalgt - - ));\n! DATA(insert OID = 815 ( \"<=\"\t PGUID 0 b t f 704 704 16 815 0 0 0 intervalle - - ));\n! DATA(insert OID = 816 ( \">=\"\t PGUID 0 b t f 704 704 16 816 0 0 0 intervalge - - ));\n \n DATA(insert OID = 843 ( \"*\"\t PGUID 0 b t f 790 700\t790 845 0 0 0 cash_mul_flt4 - - ));\n DATA(insert OID = 844 ( \"/\"\t PGUID 0 b t f 790 700\t790 0 0 0 0 cash_div_flt4 - - ));\n--- 398,409 ----\n DATA(insert OID = 808 ( \"?-\"\t PGUID 0 b t f 600 600\t 16 808 0 0 0 point_horiz - - ));\n DATA(insert OID = 809 ( \"?|\"\t PGUID 0 b t f 600 600\t 16 809 0 0 0 point_vert - - ));\n \n! DATA(insert OID = 811 ( \"=\"\t PGUID 0 b t t 704 704 16 811 812 0 0 intervaleq - - ));\n! DATA(insert OID = 812 ( \"<>\"\t PGUID 0 b t f 704 704 16 812 811 0 0 intervalne - - ));\n! DATA(insert OID = 813 ( \"<\"\t PGUID 0 b t f 704 704 16 814 816 0 0 intervallt - - ));\n! DATA(insert OID = 814 ( \">\"\t PGUID 0 b t f 704 704 16 813 815 0 0 intervalgt - - ));\n! DATA(insert OID = 815 ( \"<=\"\t PGUID 0 b t f 704 704 16 816 814 0 0 intervalle - - ));\n! DATA(insert OID = 816 ( \">=\"\t PGUID 0 b t f 704 704 16 815 813 0 0 intervalge - - ));\n \n DATA(insert OID = 843 ( \"*\"\t PGUID 0 b t f 790 700\t790 845 0 0 0 cash_mul_flt4 - - ));\n DATA(insert OID = 844 ( \"/\"\t PGUID 0 b t f 790 700\t790 0 0 0 0 cash_div_flt4 - - ));\n***************\n*** 597,603 ****\n DATA(insert OID = 1527 ( \"?-|\" PGUID 0 b t f 601 601\t16 1527 0 0 0 lseg_perp - - ));\n DATA(insert OID = 1528 ( \"?-\"\t PGUID 0 l t f 0 601\t16 1528 0 0 0 lseg_horizontal - - ));\n DATA(insert OID = 1529 ( \"?|\"\t PGUID 0 l t f 0 601\t16 1529 0 0 0 lseg_vertical - - ));\n! DATA(insert OID = 1535 ( \"=\"\t PGUID 0 b t f 601 601\t16 1535 0 0 0 lseg_eq intltsel - ));\n DATA(insert OID = 1536 ( \"#\"\t PGUID 0 b t f 601 601 600 1536 0 0 0 lseg_interpt - - ));\n DATA(insert OID = 1537 ( \"?#\"\t PGUID 0 b t f 601 628\t16 1537 0 0 0 inter_sl - - ));\n DATA(insert OID = 1538 ( \"?#\"\t PGUID 0 b t f 601 603\t16 1538 0 0 0 inter_sb - - ));\n--- 595,601 ----\n DATA(insert OID = 1527 ( \"?-|\" PGUID 0 b t f 601 601\t16 1527 0 0 0 lseg_perp - - ));\n DATA(insert OID = 1528 ( \"?-\"\t PGUID 0 l t f 0 601\t16 1528 0 0 0 lseg_horizontal - - ));\n DATA(insert OID = 1529 ( \"?|\"\t PGUID 0 l t f 0 601\t16 1529 0 0 0 lseg_vertical - - ));\n! DATA(insert OID = 1535 ( \"=\"\t PGUID 0 b t f 601 601\t16 1535 1586 0 0 lseg_eq intltsel - ));\n DATA(insert OID = 1536 ( \"#\"\t PGUID 0 b t f 601 601 600 1536 0 0 0 lseg_interpt - - ));\n DATA(insert OID = 1537 ( \"?#\"\t PGUID 0 b t f 601 628\t16 1537 0 0 0 inter_sl - - ));\n DATA(insert OID = 1538 ( \"?#\"\t PGUID 0 b t f 601 603\t16 1538 0 0 0 inter_sb - - ));\n***************\n*** 619,629 ****\n DATA(insert OID = 1578 ( \"##\"\t PGUID 0 b t f 601 601 600\t 0 0 0 0 close_lseg - - ));\n DATA(insert OID = 1585 ( \"/\"\t PGUID 0 b t f 1186 1186 1186\t 0 0 0 0 timespan_div - - ));\n \n! DATA(insert OID = 1586 ( \"<>\"\t PGUID 0 b t f 601 601\t16 1535 0 0 0 lseg_neq intltsel - ));\n! DATA(insert OID = 1587 ( \"<\"\t PGUID 0 b t f 601 601\t16 1590 0 0 0 lseg_lt intltsel - ));\n! DATA(insert OID = 1588 ( \"<=\"\t PGUID 0 b t f 601 601\t16 1589 0 0 0 lseg_le intltsel - ));\n! DATA(insert OID = 1589 ( \">\"\t PGUID 0 b t f 601 601\t16 1588 0 0 0 lseg_gt intltsel - ));\n! DATA(insert OID = 1590 ( \">=\"\t PGUID 0 b t f 601 601\t16 1587 0 0 0 lseg_ge intltsel - ));\n \n DATA(insert OID = 1591 ( \"@-@\" PGUID 0 l t f 0 601\t701 0 0 0 0 lseg_length - - ));\n \n--- 617,627 ----\n DATA(insert OID = 1578 ( \"##\"\t PGUID 0 b t f 601 601 600\t 0 0 0 0 close_lseg - - ));\n DATA(insert OID = 1585 ( \"/\"\t PGUID 0 b t f 1186 1186 1186\t 0 0 0 0 timespan_div - - ));\n \n! DATA(insert OID = 1586 ( \"<>\"\t PGUID 0 b t f 601 601\t16 1586 1535 0 0 lseg_ne intltsel - ));\n! DATA(insert OID = 1587 ( \"<\"\t PGUID 0 b t f 601 601\t16 1589 1590 0 0 lseg_lt intltsel - ));\n! DATA(insert OID = 1588 ( \"<=\"\t PGUID 0 b t f 601 601\t16 1590 1589 0 0 lseg_le intltsel - ));\n! DATA(insert OID = 1589 ( \">\"\t PGUID 0 b t f 601 601\t16 1587 1588 0 0 lseg_gt intltsel - ));\n! DATA(insert OID = 1590 ( \">=\"\t PGUID 0 b t f 601 601\t16 1588 1587 0 0 lseg_ge intltsel - ));\n \n DATA(insert OID = 1591 ( \"@-@\" PGUID 0 l t f 0 601\t701 0 0 0 0 lseg_length - - ));\n \n\n\n",
"msg_date": "Wed, 28 Oct 1998 18:51:28 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
},
{
"msg_contents": "On Wed, 28 Oct 1998, Tom Lane wrote:\n\n> Actually, the darn thing is *crawling* with bugs.\n\nGuess BETA3 won't be the release, eh? *grin*\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 28 Oct 1998 20:48:34 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
},
{
"msg_contents": "Tom Lane <[email protected]> writes:\n> This was, um, fruitful.\n\nMy *goodness*, that was a good idea.\n\nI have now located and repaired ninety-three distinct bugs in\npg_operator.h, all of the form \"operator A has an incorrect com, negate,\nor sort link to operator B\". Almost none of them required any semantic\nanalysis to spot --- I found them by looking for conditions like A links\nto B but B doesn't link to A, or A claims B is its commutation but B\ndoesn't have the right input data types to be that, etc.\n\nThe tinterval regress test now produces believable results --- ie,\nthe order in which allegedly-sorted intervals appear actually agrees\nwith the length of the intervals. I would imagine that we can get\nrid of the special-case NetBSD expected output for tinterval, as well.\n\nI will shortly commit these changes (as soon as I finish another\nbuild/regress pass). I will also commit a new regression test script\nthat looks for all the test conditions that I used to locate these\nproblems, in hopes that no new bugs of this ilk will creep in.\n\nThere is one unfixed bug in pg_operator, which I let go because I didn't\nwant to make the decision as to what to do with it, but it needs to be\nfixed. Namely, there is a conflict between OIDs 512 (on_ppath) and 754\n(pt_contained_path), both of which claim to be the implementation of\n\"point @ path\":\n\nDATA(insert OID = 512 ( \"@\"\t PGUID 0 b t f 600 602 16 0 0 0 0 on_ppath intltsel intltjoinsel ));\n\nDATA(insert OID = 754 ( \"@\"\t PGUID 0 b t f 600 602\t 16 755 0 0 0 pt_contained_path - - ));\n\nI imagine that which one you actually get is a quasi-random matter...\nwe need to decide which one is the One True @, and change the oprname\nof the other one.\n\nI suggest that it'd be a Good Idea to develop similar check scripts for\nthe other hand-maintained tables. But I figure I've done enough for one\nday...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 28 Oct 1998 21:53:49 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
},
{
"msg_contents": "Tom Lane <[email protected]> writes:\n> > This was, um, fruitful.\n> \n> My *goodness*, that was a good idea.\n> \n> I have now located and repaired ninety-three distinct bugs in\n> pg_operator.h, all of the form \"operator A has an incorrect com, negate,\n> or sort link to operator B\". Almost none of them required any semantic\n> analysis to spot --- I found them by looking for conditions like A links\n> to B but B doesn't link to A, or A claims B is its commutation but B\n> doesn't have the right input data types to be that, etc.\n\nBravo! I did this for Illustra a couple years ago. What a pain. But it is\nnice to get right answers. Also, it lets the planner generate better plans.\nIf there is a negator or commutator it can use it instead of generating extra\nsteps.\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - If simplicity worked, the world would be overrun with insects. -\n",
"msg_date": "Wed, 28 Oct 1998 19:05:29 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> Actually, the darn thing is *crawling* with bugs.\n\nOK, sit down and take a deep breath. And another one. OK, do you still\nsee bugs everywhere? Oh, that's too bad. We'll send someone over right\naway. Stay away from sharp objects, OK? *grin*\n\n> The comparison operators on booleans, tintervals, and lsegs are all\n> pretty seriously busted --- they have incorrect commute and/or negate\n> links. I am worried about what other similar bugs may be in there,\n> but short of groveling over the entries by hand, I don't know what \n> else to do to check the table further.\n\nWell, we should grovel over the entries by hand, for v6.5. I'd love to\nget that thing cleaned up. But I'll bet that some of what you're\nfinding, although inconsistant, has no adverse effect currently. And it\nisn't worse than in previous releases (and in some places it is better). \n\nIn particular, the commute and negate links are not used as extensively\nas you might think they should be. But I'd like them to be used more in\nthe future.\n\nI vaguely recall that I do (mis-)use the commute operators for entries\nfor which the commute operator would not otherwise be meaningful to help\nreorder expressions in the parser. I think that for mixed-type operators\nI use the commute column to suggest an alternate operator when trying to\nfind candidates in the new type coersion stuff.\n\n> BTW Bruce's fix of this morning wasn't right either --- he misspelled\n> the name of the lseg_ne comparison routine.\n\nWell, fix that for sure...\n\n> I am currently running a build and regression test for the attached\n> diffs, which I propose to commit if the regress passes.\n> I expect to find that this explains the tinterval regression \n> differences I've been seeing. The tinterval expected file is clearly \n> wrong, because it contains \"sorted\" lists of tintervals that are \n> obviously not in order.\n\ntinterval is an \"orphan\" type. If it is really a useful type for people,\nwe should adopt it and fix it up.\n\n> BTW, is it sufficient to edit pg_operator.h, rebuild, reinstall, and\n> initdb? Or is there some additional incantation that needs to be made\n> when patching pg_operator.h?\n\nNot that I can recall. You shouldn't reassign OIDs, and there is a small\nchance that an OID you remove is actually referenced in one of the\nindexing tables (I don't recall the exact layout). I've found that I\nneed to do a full \"make clean install\", otherwise some things end up in\nan inconsistant state, probably due to incomplete Makefile dependencies.\n\n> Stop the presses .... looks like int8 and int4-vs-int8 comparators \n> are bogus too ... geez ...\n\nIs this a robust system or what! You can get half the tables wrong and\nstill get good results. Those AI guys are green with envy :)\n\nbtw, what'd I do wrong on those?\n\n - Tom\n",
"msg_date": "Thu, 29 Oct 1998 03:54:06 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> As long as we're doing last-minute changes that force initdb ...\n> \n> I believe I've just identified another error in pg_operator.h.\n> Specifically, this tuple is bogus and ought to be deleted:\n> \n> DATA(insert OID = 619 ( \"<\"\t PGUID 0 b t f 704 704 16 0 0 0 0 intervalct - - ));\n\nYep. I have removed the entry. It didn't even look like it belonged\nthere.\n\n> \n> This tuple defines \"<\" on tinterval's (type 704 = TINTERVAL) as\n> being \"intervalct\", or interval-contained-in. Now that doesn't\n> make a lot of sense. Furthermore there is already a tuple\n> that defines \"<<\" as being intervalct:\n> \n> DATA(insert OID = 573 ( \"<<\"\t PGUID 0 b t f 704 704 16 0 0 0 0 intervalct - - ));\n> \n> and there is another tuple that defines \"<\" on tintervals as\n> being \"intervallt\":\n> \n> DATA(insert OID = 813 ( \"<\"\t PGUID 0 b t f 704 704 16 813 0 0 0 intervallt - - ));\n> \n> I found this while trying to track down why the tinterval regression\n> test fails on HPUX (it produces the right tuples, but in the wrong\n> sort order ... and unlike the rules case, the select does have an\n> ORDER BY clause). As far as I can tell from the CVS logs, the bogosity\n> has been in there for at least a year.\n\nThat is an amazing way to find a bad pg_operator entry.\n\n> \n> It looks to me like the bogus tuple is masking the correct definition\n> of \"<\" and causing tintervals to sort incorrectly. Why this doesn't\n> happen on many other machines isn't clear, but maybe it has something\n> to do with the platform-dependency of qsort that we already identified?\n> \n> Anyway, I'm a rank newbie at hacking this stuff, so I'm not about\n> to commit this change without asking someone else to check my work.\n> But if it is indeed wrong, right now seems like a good time to fix it...\n\nI have done the dirty deed, so Marc can break my fingers.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 23:08:13 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> On Wed, 28 Oct 1998, Tom Lane wrote:\n> \n> > As long as we're doing last-minute changes that force initdb ...\n> \n> You guys all know I'm going to be a super-prick next release, right?\n> *grin* Just a pre-warning... :)\n\nMarc is a Canadian Frenchy. If they are want to succeed from their own\ncountry, they could do anything.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 23:09:28 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> On Wed, 28 Oct 1998, Tom Lane wrote:\n> \n> > Actually, the darn thing is *crawling* with bugs.\n> \n> Guess BETA3 won't be the release, eh? *grin*\n\nI can see an argument for letting it go for 6.5, but if it is broken,\nand is showing up in regression testing, I think we have to fix it.\n\nWe can't fix it in 6.4.*. No initdb for that.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 23:20:09 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "On Wed, 28 Oct 1998, Bruce Momjian wrote:\n\n> I have done the dirty deed, so Marc can break my fingers.\n\n\tOne of these days, I'll brave crossing the border into \"red neck\ncop territory\" :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 00:35:34 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> Tom Lane <[email protected]> writes:\n> > This was, um, fruitful.\n> \n> My *goodness*, that was a good idea.\n> \n> I have now located and repaired ninety-three distinct bugs in\n> pg_operator.h, all of the form \"operator A has an incorrect com, negate,\n> or sort link to operator B\". Almost none of them required any semantic\n> analysis to spot --- I found them by looking for conditions like A links\n> to B but B doesn't link to A, or A claims B is its commutation but B\n> doesn't have the right input data types to be that, etc.\n\nGreen light. Go.\n\n> \n> The tinterval regress test now produces believable results --- ie,\n> the order in which allegedly-sorted intervals appear actually agrees\n> with the length of the intervals. I would imagine that we can get\n> rid of the special-case NetBSD expected output for tinterval, as well.\n> \n> I will shortly commit these changes (as soon as I finish another\n> build/regress pass). I will also commit a new regression test script\n> that looks for all the test conditions that I used to locate these\n> problems, in hopes that no new bugs of this ilk will creep in.\n\nWhat I did was to make a file in the include/catalog directory called\ntemplate1_check.sql and pg_attribute_check.sql. These are SQL\nstatements that check various catalogs and report problems where things\nare missing joins. Perhaps we could put something in there, or move\nthese to the regression directory and include them in your stuff. They\nare not platform-specific.\n\nWould be nice so we could spot the problems right away. Or perhaps put\na script in include/catalogs that checks all the system tables, and run\nthat from the regression directory. That would be nice too.\n\n> \n> There is one unfixed bug in pg_operator, which I let go because I didn't\n> want to make the decision as to what to do with it, but it needs to be\n> fixed. Namely, there is a conflict between OIDs 512 (on_ppath) and 754\n> (pt_contained_path), both of which claim to be the implementation of\n> \"point @ path\":\n> \n> DATA(insert OID = 512 ( \"@\"\t PGUID 0 b t f 600 602 16 0 0 0 0 on_ppath intltsel intltjoinsel ));\n> \n> DATA(insert OID = 754 ( \"@\"\t PGUID 0 b t f 600 602\t 16 755 0 0 0 pt_contained_path - - ));\n\n\nI see \"@\" means \"on\" sometimes, and \"contained\" sometimes, and they use\n\"@\" for both uses for the point/path combination. Looks like a problem\nthat \"@\" applies to point/path, and \"on\" and \"contained\" are both valid.\nHowever, they seem to mean the same thing. Are on_ppath and\npt_contained_path doing the same thing. Thomas could help here.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Wed, 28 Oct 1998 23:36:12 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "On Wed, 28 Oct 1998, Bruce Momjian wrote:\n\n> > On Wed, 28 Oct 1998, Tom Lane wrote:\n> > \n> > > As long as we're doing last-minute changes that force initdb ...\n> > \n> > You guys all know I'm going to be a super-prick next release, right?\n> > *grin* Just a pre-warning... :)\n> \n> Marc is a Canadian Frenchy. If they are want to succeed from their own\n> country, they could do anything.\n\n\tHey hey, careful there. I'm against seperatism, for one. And I\nwas born in Nova Scotia for two...just got alot of Frenchy blood in me is\nall :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 00:36:22 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "On Wed, 28 Oct 1998, Bruce Momjian wrote:\n\n> > On Wed, 28 Oct 1998, Tom Lane wrote:\n> > \n> > > Actually, the darn thing is *crawling* with bugs.\n> > \n> > Guess BETA3 won't be the release, eh? *grin*\n> \n> I can see an argument for letting it go for 6.5, but if it is broken,\n> and is showing up in regression testing, I think we have to fix it.\n> \n> We can't fix it in 6.4.*. No initdb for that.\n\n\tBruce, can you and Thomas go through Tom's patch with a fine\ntoothed comb before its committed? this close to release (5days!), I'd\nrather have a tri-agreement on it, but from what David mentioned, it seems\nthat it would be \"a good thing\" to get in before v6.4 is released...\n\n\tIf all three of you feel it is a) right and b) safe, throw it in.\nI'll do up a quick BETA4 (release candidate 2) as soon as you tell me\nwhether or not its being/been committed...\n\n Marc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 00:39:29 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> \tBruce, can you and Thomas go through Tom's patch with a fine\n> toothed comb before its committed? this close to release (5days!), I'd\n> rather have a tri-agreement on it, but from what David mentioned, it seems\n> that it would be \"a good thing\" to get in before v6.4 is released...\n> \n> \tIf all three of you feel it is a) right and b) safe, throw it in.\n> I'll do up a quick BETA4 (release candidate 2) as soon as you tell me\n> whether or not its being/been committed...\n\nLooks like it is already in. I will check it. Thomas is checking the\n'@' issue right now on IRC.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 29 Oct 1998 00:08:01 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> > Marc is a Canadian Frenchy. If they are want to succeed from their own\n> > country, they could do anything.\n> \n> \tHey hey, careful there. I'm against seperatism, for one. And I\n> was born in Nova Scotia for two...just got alot of Frenchy blood in me is\n> all :)\n\nI just wanted to use the word \"Frenchy\", that's all.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 29 Oct 1998 00:10:30 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "On Thu, 29 Oct 1998, Bruce Momjian wrote:\n\n> > > Marc is a Canadian Frenchy. If they are want to succeed from their own\n> > > country, they could do anything.\n> > \n> > \tHey hey, careful there. I'm against seperatism, for one. And I\n> > was born in Nova Scotia for two...just got alot of Frenchy blood in me is\n> > all :)\n> \n> I just wanted to use the word \"Frenchy\", that's all.\n\n\t*roll eyes* :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 01:17:23 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> Bruce, can you and Thomas go through Tom's patch with a fine\n> toothed comb before its committed? this close to release (5days!)\n\nI'm sure they are good things, and Tom Lane has been careful and\nthoughtful about what he does in past patches (and sticks around to pick\nup the pieces :). Does the regression test still behave in all other\nrespects Tom?\n\nI'm a bit worried about keeping up with the docs schedule. So if I look\nat these tables, and if my docs get behind, can you consider giving me a\ncouple of days to recover (only if I need it)? I'm out of town this\nweekend too, which doesn't help...\n\n - Tom\n",
"msg_date": "Thu, 29 Oct 1998 05:26:07 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "[email protected] (David Gould) writes:\n> Also, it lets the planner generate\n> better plans. If there is a negator or commutator it can use it\n> instead of generating extra steps.\n\nWell, I did *not* go looking for links that should be there and weren't\n(except in the very special case that the reverse link existed). I\njust tried to sanity-check the existing links.\n\nI agree that it would be nice to look for missing links that should\nbe added ... but that is a performance enhancement, not a bug fix,\nso I am not eager to do it this close to release. We should do\nanother pass over this table after 6.4 is out the door.\n\n(Another reason I didn't try to do that is that I've got no good\nidea how to find missing links, short of brain-numbingly tedious\nhand examination. Can you suggest any automated way of finding\nmissing links, or at least finding likely things to look at?)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 29 Oct 1998 00:36:31 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
},
{
"msg_contents": "On Thu, 29 Oct 1998, Thomas G. Lockhart wrote:\n\n> > Bruce, can you and Thomas go through Tom's patch with a fine\n> > toothed comb before its committed? this close to release (5days!)\n> \n> I'm sure they are good things, and Tom Lane has been careful and\n> thoughtful about what he does in past patches (and sticks around to pick\n> up the pieces :). Does the regression test still behave in all other\n> respects Tom?\n\n\tQuite frankly, this close to the release, I'd kinda ask the same\nif it were any of the other committers too...I'm not really singling out\nTom :)\n\n> I'm a bit worried about keeping up with the docs schedule. So if I look\n> at these tables, and if my docs get behind, can you consider giving me a\n> couple of days to recover (only if I need it)? I'm out of town this\n> weekend too, which doesn't help...\n\n\tConsider it done...if I hear nothing from you, I'll still consider\nMonday a go...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 01:36:53 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n>> Tom Lane <[email protected]> writes:\n>> I will also commit a new regression test script\n>> that looks for all the test conditions that I used to locate these\n>> problems, in hopes that no new bugs of this ilk will creep in.\n\n> What I did was to make a file in the include/catalog directory called\n> template1_check.sql and pg_attribute_check.sql. These are SQL\n> statements that check various catalogs and report problems where things\n> are missing joins. Perhaps we could put something in there, or move\n> these to the regression directory and include them in your stuff.\n\nI saw those but it wasn't clear to me when they would get applied or\nwhether they were hand-generated or derived from something else. So\nI went and made a new regression test, because I think I comprehend\nthose. If you want to fold the opr_sanity regress test into one of\nthe sql files in include/catalog, go right ahead. (Or, maybe those\nfiles should be pushed over to regression testing? I dunno.)\n\n> I see \"@\" means \"on\" sometimes, and \"contained\" sometimes, and they use\n> \"@\" for both uses for the point/path combination. Looks like a problem\n> that \"@\" applies to point/path, and \"on\" and \"contained\" are both valid.\n> However, they seem to mean the same thing. Are on_ppath and\n> pt_contained_path doing the same thing. Thomas could help here.\n\nI would expect that on_ppath checks to see if the point is on (touches)\nthe path, whereas pt_contained_path checks to see if the point is within\nthe area enclosed by the path. But I haven't looked to see if that's\nwhat the author of the code thought...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 29 Oct 1998 00:47:44 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
},
{
"msg_contents": "> [email protected] (David Gould) writes:\n> > Also, it lets the planner generate\n> > better plans. If there is a negator or commutator it can use it\n> > instead of generating extra steps.\n> \n> Well, I did *not* go looking for links that should be there and weren't\n> (except in the very special case that the reverse link existed). I\n> just tried to sanity-check the existing links.\n> \n> I agree that it would be nice to look for missing links that should\n> be added ... but that is a performance enhancement, not a bug fix,\n> so I am not eager to do it this close to release. We should do\n> another pass over this table after 6.4 is out the door.\n> \n> (Another reason I didn't try to do that is that I've got no good\n> idea how to find missing links, short of brain-numbingly tedious\n> hand examination. Can you suggest any automated way of finding\n> missing links, or at least finding likely things to look at?)\n\nYes. See include/catalog/template1_check.sql. Uses SQL to do checks.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 29 Oct 1998 01:13:26 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> \n> [email protected] (David Gould) writes:\n> > Also, it lets the planner generate\n> > better plans. If there is a negator or commutator it can use it\n> > instead of generating extra steps.\n> \n> Well, I did *not* go looking for links that should be there and weren't\n> (except in the very special case that the reverse link existed). I\n> just tried to sanity-check the existing links.\n> \n> I agree that it would be nice to look for missing links that should\n> be added ... but that is a performance enhancement, not a bug fix,\n> so I am not eager to do it this close to release. We should do\n> another pass over this table after 6.4 is out the door.\n> \n> (Another reason I didn't try to do that is that I've got no good\n> idea how to find missing links, short of brain-numbingly tedious\n> hand examination. Can you suggest any automated way of finding\n> missing links, or at least finding likely things to look at?)\n> \n> \t\t\tregards, tom lane\n\nWell, I have to admit, I did the brain-numbingly tedious hand examination.\nBut, mostly it is straight forward. If you have op 'lt' for type foo as\nfunction foolt, then it probably has negetor 'ge' as function 'fooge'.\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - If simplicity worked, the world would be overrun with insects. -\n",
"msg_date": "Wed, 28 Oct 1998 23:15:41 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> Yes. See include/catalog/template1_check.sql. Uses SQL to do checks.\n\nHmm. OK, that script has some overlap with what I did last night.\nBut is it run automatically? I think putting the checks into a\nregression test is a better plan.\n\nAlso, in playing around with template1_check.sql, I discovered that\npg_operator OID 644\n\nDATA(insert OID = 644 ( \"<>\"\t PGUID 0 b t f 30 30 16 644 649 0 0 oid8ne neqsel neqjoinsel ));\n\npoints to a nonexistent operator --- there is no oid8ne in pg_proc, nor\nin the code... anyone feel like fixing it?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 29 Oct 1998 11:52:46 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
},
{
"msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > Yes. See include/catalog/template1_check.sql. Uses SQL to do checks.\n> \n> Hmm. OK, that script has some overlap with what I did last night.\n> But is it run automatically? I think putting the checks into a\n> regression test is a better plan.\n\nYou were right, template1_check.sql is generated by\npgsql/contrib/findoidjoins/make_oidjoin_check.\n\nIt is not run automatically, but should be, just as duplicate_oids\nshould be.\n\n> \n> Also, in playing around with template1_check.sql, I discovered that\n> pg_operator OID 644\n> \n> DATA(insert OID = 644 ( \"<>\"\t PGUID 0 b t f 30 30 16 644 649 0 0 oid8ne neqsel neqjoinsel ));\n> \n> points to a nonexistent operator --- there is no oid8ne in pg_proc, nor\n> in the code... anyone feel like fixing it?\n\nAdded.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Thu, 29 Oct 1998 13:02:43 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "> points to a nonexistent operator --- there is no oid8ne in pg_proc, \n> nor in the code... anyone feel like fixing it?\n\nYou de man! I just committed inet regression test fixes and resolved the\n\"@\" operator problem. I'm back to working on docs for now...\n\n - Tom\n\nactually, at the moment I'm back to going to work :)\n",
"msg_date": "Thu, 29 Oct 1998 18:20:10 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> I just committed inet regression test fixes and resolved the\n> \"@\" operator problem.\n\nGreat! I think that means it's safe to put out a BETA4, Marc.\n\nLeastwise I promise to stop breaking things for a while ;-)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 29 Oct 1998 13:27:47 -0500",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
},
{
"msg_contents": "On Thu, 29 Oct 1998, Tom Lane wrote:\n\n> \"Thomas G. Lockhart\" <[email protected]> writes:\n> > I just committed inet regression test fixes and resolved the\n> > \"@\" operator problem.\n> \n> Great! I think that means it's safe to put out a BETA4, Marc.\n\n\tDone, BETA4 is rolled...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 22:21:47 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Another bug in pg_operator.h "
}
] |
[
{
"msg_contents": "Hello!\n\nI'm the original reporter of \"vacuum crash\" problem.\nThank you for your quick fix.\n\nNow, I found another problem in nbtsearch.c.\nThis is my patch for it.\n# But, what I want to know is why postgres reaches this FATAL status...\n\n========================================================================\n*** nbtsearch.c.orig\tTue Oct 20 16:42:59 1998\n--- nbtsearch.c\tTue Oct 20 16:43:15 1998\n***************\n*** 217,223 ****\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, P_FIRSTKEY),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n \t\t\t\t\t\t\telog(FATAL, \"btree: BTP_CHAIN flag was expected in %s (access = %s)\",\n! \t\t\t\t\t\t\t\t rel->rd_rel->relname, access ? \"bt_write\" : \"bt_read\");\n \t\t\t\t\t\tif (_bt_skeycmp(rel, keysz, scankey, page,\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, offmax),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n--- 217,223 ----\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, P_FIRSTKEY),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n \t\t\t\t\t\t\telog(FATAL, \"btree: BTP_CHAIN flag was expected in %s (access = %s)\",\n! \t\t\t\t\t\t\t\t rel->rd_rel->relname.data, access ? \"bt_write\" : \"bt_read\");\n \t\t\t\t\t\tif (_bt_skeycmp(rel, keysz, scankey, page,\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, offmax),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n========================================================================\n\nAnd with the following script, you can find suspicious lines which\nhave pointer type incompatibility but work *fortunately* on most\nsystems.\n\n========================================================================\n#/bin/sh\n\nfind PGSRC -type f -name '*.[ch]' -print |\nxargs egrep -e 'rd_rel->relname' /dev/null |\negrep -v 'relname\\)?\\.data'\n========================================================================\n\nRegards\n\n-- \nASCII CORPORATION\nTechnical Center\nSHIOZAKI Takehiko\n<[email protected]>\n",
"msg_date": "Thu, 29 Oct 1998 10:48:23 +0900 (JST)",
"msg_from": "SHIOZAKI Takehiko <[email protected]>",
"msg_from_op": true,
"msg_subject": "relname.data"
}
] |
[
{
"msg_contents": "Hi.\n\nI have not followed hackers regularly, so im sure this has probably been \nasked before. Anyways, I tried the 6.4 beta 3 today, and noticed that\nthe\nword \"USER\" has been reserved now. Since I have an app that uses a\ncolumn\nnamed \"user\" I went into backend/parser/gram.y and added a line like:\n\n | USER { $$ = \"user\" }\n\ninto that file so that USER would be accepted.\n\nHowever, I noticed that this breaks a lot of queries :). Apparently\ndoing a\n\"SELECT user\" now returns your username. So in order to make my queries\nwork\nI would have to do \"SELECT table.user FROM table....\". If there is no\nother\nway around it, Ill probably just have to rename the colunm and track\ndown all\nreferences to it in my application (ugh :(). \n\nSO my question is this: Is there anything else that could be done to\nmake\nthis work? Or will we break the \"SELECT USER\" usefulness by doing\nsomething \nlike this? Im guessing its probably not going to be possible :) But\nthough\nI would ask those a little more familiar with the source tree than me\nsince\nI just started looking at at now :)\n\nMike\n",
"msg_date": "Wed, 28 Oct 1998 21:51:48 -0600",
"msg_from": "Michael J Schout <[email protected]>",
"msg_from_op": true,
"msg_subject": "6.4 and reserved word USER..."
},
{
"msg_contents": "\nDidn't we make that change for v6.3.2? *raised eyebrow*\n\nOn Wed, 28 Oct 1998, Michael J Schout wrote:\n\n> Hi.\n> \n> I have not followed hackers regularly, so im sure this has probably been \n> asked before. Anyways, I tried the 6.4 beta 3 today, and noticed that\n> the\n> word \"USER\" has been reserved now. Since I have an app that uses a\n> column\n> named \"user\" I went into backend/parser/gram.y and added a line like:\n> \n> | USER { $$ = \"user\" }\n> \n> into that file so that USER would be accepted.\n> \n> However, I noticed that this breaks a lot of queries :). Apparently\n> doing a\n> \"SELECT user\" now returns your username. So in order to make my queries\n> work\n> I would have to do \"SELECT table.user FROM table....\". If there is no\n> other\n> way around it, Ill probably just have to rename the colunm and track\n> down all\n> references to it in my application (ugh :(). \n> \n> SO my question is this: Is there anything else that could be done to\n> make\n> this work? Or will we break the \"SELECT USER\" usefulness by doing\n> something \n> like this? Im guessing its probably not going to be possible :) But\n> though\n> I would ask those a little more familiar with the source tree than me\n> since\n> I just started looking at at now :)\n> \n> Mike\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 29 Oct 1998 00:37:22 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] 6.4 and reserved word USER..."
},
{
"msg_contents": "> > the word \"USER\" has been reserved now.\n> > SO my question is this: Is there anything else that could be done \n> > to make this work?\n\nNot that this helps you much, \"USER\" is an SQL92 reserved word. There is\na new chapter in the User's Guide on syntax which shows lists of the\nreserved and unreserved words in SQL92, SQL3, and/or Postgres to help\npeople avoid trouble.\n\nYou can do your queries like\n\n SELECT \"user\" from table...\n\nto keep using that column name, but it would still require you to find\nand change all of the instances in your apps. I assume that adding the\nColId definition of USER in gram.y resulted in massive numbers of\nshift/reduce conflicts? If not, then we should just put it in there, but\nI can't see why you wouldn't get the shift/reduce problems since it\nwould be allowed the same places as CURRENT_USER/USER are.\n\n - Tom\n",
"msg_date": "Thu, 29 Oct 1998 05:33:00 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] 6.4 and reserved word USER..."
}
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.