Обсуждение: v7.0 bug report: problems on Linux/Alpha
Hi all,
there appears to be some problems with 64-bit-cleaness of
v7.0. I've got it the identical operations working fine on both Linux/x86
and OpenBSD/x86, but it dies on my Alpha box.
Alpha specs: AS200 4/166 w/72MB running Debian Linux 2.1 with kernel 2.2.15
Extra ./configure options: none
Compiler used: egcs-1.1.2
Libraries linked to:
libc6.1 2.0.7.19981211 GNU C Library: shared libraries
libncurses4 4.2-3 Shared libraries for terminal handling
libreadlineg2 2.1-12 GNU readline and history libraries
Admittedly, some of these are getting ancient, but due to it working on
the x86 Linux box with roughly the same library versions, I don't think
that's the problem.
Problem description:
--------------------
"\d tablename", "\do", "\df" and perhaps other commands inside psql
produce the following error: "getattproperties: no attribute tuple XXXX -2"
The numeric XXXX value is different depending on the command issued, but
is always the same value for the same command. The problem has continued
even after another initdb.
It doesn't seem to be related to user table layouts - something to do with
OID processing according to my fiddling below.
Example:
db=> \d
List of relations
Name | Type | Owner
----------------+----------+--------
author | table | planet
db=> \d author
********* QUERY *********
SELECT relhasindex, relkind, relchecks, reltriggers, relhasrules
FROM pg_class WHERE relname='author'
*************************
********* QUERY *********
SELECT a.attname, t.typname, a.attlen, a.atttypmod, a.attnotnull, a.atthasdef, a
.attnum
FROM pg_class c, pg_attribute a, pg_type t
WHERE c.relname = 'author'
AND a.attnum > 0 AND a.attrelid = c.oid AND a.atttypid = t.oid
ORDER BY a.attnum
*************************
ERROR: getattproperties: no attribute tuple 1259 -2
db=> select oid from pg_class where oid = 1259;
ERROR: getattproperties: no attribute tuple 1259 -2
db=> \do
ERROR: getattproperties: no attribute tuple 1255 -2
db=> \df
ERROR: getattproperties: no attribute tuple 1247 -2
I have no idea where to start with tracking this down, or I'd be off
quietly bugfinding :-)
Hope this helps someone find and fix the problem - it means pg_dump won't
work, which is far more serious than not being able to describe a table...
Regards,
Nic.
-- Nic Bellamy <nic@bellamy.co.nz>
Director, Bellamy Consulting Ltd.
> there appears to be some problems with 64-bit-cleaness of
> v7.0. I've got it the identical operations working fine on both Linux/x86
> and OpenBSD/x86, but it dies on my Alpha box.
> I have no idea where to start with tracking this down, or I'd be off
> quietly bugfinding :-)
There are patches available to get Postgres working on Alpha boxes. We
hope to have some parts of Postgres rewritten in the next version
which should make patching unnecessary, but at the moment patching was
the easiest and cleanest way to get Alpha support.
See the hackers list archives for the patches, or post a message to
the hackers list asking about them. Ryan Kirkpatrick (sp?) has been
working on the patches, so the patches might be posted under his name.
Good luck.
- Thomas
--
Thomas Lockhart lockhart@alumni.caltech.edu
South Pasadena, California
On Wed, 17 May 2000, Thomas Lockhart wrote: > There are patches available to get Postgres working on Alpha boxes. We > hope to have some parts of Postgres rewritten in the next version > which should make patching unnecessary, but at the moment patching was > the easiest and cleanest way to get Alpha support. > > See the hackers list archives for the patches, or post a message to > the hackers list asking about them. Ryan Kirkpatrick (sp?) has been > working on the patches, so the patches might be posted under his name. Thanks to the name dropping, I managed to find the patch on Ryan K.'s page (http://www.rkirkpat.net/), and it has fixed the problems I was having. Many thanks for the pointer :-) There were a couple of rejected sections in the patch, but they were easily applied by hand. If wanted, I can chuck a copy of the diff vs. 7.0 on an ftp site, or post it to this list. Regards, Nic. -- Nic Bellamy <nic@bellamy.co.nz> Director, Bellamy Consulting Ltd.
> There were a couple of rejected sections in the patch, but they were
> easily applied by hand. If wanted, I can chuck a copy of the diff vs. 7.0
> on an ftp site, or post it to this list.
Ryan has graduated college (in top of class, btw) and and claims to
have lots of free time coming up to work on Alpha patches. Y'all may
want to coordinate on generating and testing some patches for 7.0. But
a good starting point would be to post some patches from your tree.
Tom Lane is planning a fmgr rewrite that will (hopefully) make these
patches obsolete and unnecessary, but that won't be available until
the next full release in a few months. So solid patches will be a
benefit in the meantime.
Regards.
- Thomas
--
Thomas Lockhart lockhart@alumni.caltech.edu
South Pasadena, California
On Thu, 18 May 2000, Thomas Lockhart wrote:
> Ryan has graduated college (in top of class, btw) and and claims to
> have lots of free time coming up to work on Alpha patches. Y'all may
> want to coordinate on generating and testing some patches for 7.0. But
> a good starting point would be to post some patches from your tree.
I'm happy to help test, although my databases tend to be rather simple.
I've put the patch up at:
ftp://ftp.wibble.net/pub/stuff/postgresql-7.0-alpha.patch.gz
It's identical to RKs beta5 patch, but cleaned up to apply properly to
the 7.0 release.
Regards,
Nic.
-- Nic Bellamy <nic@bellamy.co.nz>
Director, Bellamy Consulting Ltd.
> I've put the patch up at:
> ftp://ftp.wibble.net/pub/stuff/postgresql-7.0-alpha.patch.gz
> It's identical to RKs beta5 patch, but cleaned up to apply properly to
> the 7.0 release.
Lamar, is that enough to get you going on a .src.rpm supporting alpha?
Someone can test a binary after that...
- Thomas
--
Thomas Lockhart lockhart@alumni.caltech.edu
South Pasadena, California
Thomas Lockhart wrote: > > > I've put the patch up at: > > ftp://ftp.wibble.net/pub/stuff/postgresql-7.0-alpha.patch.gz > > It's identical to RKs beta5 patch, but cleaned up to apply properly to > > the 7.0 release. > > Lamar, is that enough to get you going on a .src.rpm supporting alpha? > Someone can test a binary after that... Yes, it is. -2 RPMset to be built within next 48 hours. -- Lamar Owen WGCR Internet Radio 1 Peter 4:11