Re: BUG #3763: crash after create table with primary keydefined

admin@abp.pl wrote:
> Dnia Wt Listopada 20 2007, 12:46, Heikki Linnakangas napisa³(a):
>> Simon Riggs wrote:
>>> On Tue, 2007-11-20 at 10:12 +0000, migo wrote:
>>>> =>  CREATE TABLE phpbb_acl_roles_data ( role_id INT4 DEFAULT '0' NOT NULL
>>>> CHECK (role_id >= 0), auth_option_id INT4 DEFAULT '0' NOT NULL CHECK
> (auth_option_id >= 0), auth_setting INT2 DEFAULT '0' NOT NULL, PRIMARY
> KEY
>>>> (role_id, auth_option_id) );
>>>> server closed the connection unexpectedly
>>>>         This probably means the server terminated abnormally
>>>>         before or while processing the request.
>>>> The connection to the server was lost. Attempting reset: Succeeded.
> wolni=>
>>>> In logs after this:
>>>> Nov 20 11:02:04 www postgres[32757]: [4-1] wolni>NOTICE:  CREATE TABLE /
>>>> PRIMARY KEY will create implicit index "phpbb_acl_roles_data_pkey" for
> table
>>>> Nov 20 11:02:04 www postgres[32757]: [4-2]  "phpbb_acl_roles_data" Nov
> 20 11:02:04 www postgres[32758]: [2-1] [unknown]>LOG:  connection
> received: host=[local]
>>>> Nov 20 11:02:04 www postgres[32758]: [3-1] wolni>LOG:  connection
> authorized: user=wolni database=wolni
>>>> Nov 20 11:02:04 www postgres[32731]: [2-1] >LOG:  server process (PID
> 32757)
>>>> was terminated by signal 11
>>>> Nov 20 11:02:04 www postgres[32731]: [3-1] >LOG:  terminating any
> other
>>>> active server processes
>>>> Nov 20 11:02:04 www postgres[32758]: [4-1] wolni>WARNING:  terminating
> connection because of crash of another server process
>>> Looks to me that the process that crashed was not the process that
> issued the CREATE TABLE command. So it could be something else
> entirely.
>
> Strange thinks it that this happen only when i log into psql via "wolni"
> user. When i do this as postgres admin all works fint too.
> User 'wolni' has granted all privilages to database 'wolni'.
>
>> To me it does look like it was the CREATE TABLE that crashed. The NOTICE
>>   shows that it's process ID 32757 that's running the CREATE TABLE, and
>> the LOG line later on says that that's the process that crashed.
>>
>>> Is this behaviour repeatable?
>> I couldn't reproduce this on my laptop with 8.2.4. Migo, is it possible
> for you to get a core dump and a back trace out of it? Is there anything
> unusual about the system / compiler / compiler options used?
>
> Its gentoo distribution.
>
> # gcc -v
> gcc version 3.4.6 (Gentoo Hardened 3.4.6-r2 p1.5, ssp-3.4.6-1.0,
> pie-8.7.10)
>
> with opts:
> CFLAGS="-march=prescott -O2 -pipe -fforce-addr -fomit-frame-pointer"
> CXXFLAGS="${CFLAGS}"
> MAKEOPTS="-j2"

Ok, you've used -fomit-frame-pointer, so we can't get meaningful debug
information out of it. :-(

Can you please recompile without -fomit-frame-pointer, and reconfigure
postgres with --enable-debug and --enable-cassert? You don't want the
--enable-cassert in production, since it slows things down, but
--enable-debug shouldn't hurt.

> On another my system (with gentoo too, and postgresql 8.2.4 all queries
> works).

Is it on similar hardware? Same configure and compiler options?

--
   Heikki Linnakangas
   EnterpriseDB   http://www.enterprisedb.com