Thread: ANALYSE on Solaris 8 bug
Patrick Vachon kaleidelf AT hotmail.com SPARC Sun Blade 100 Solaris 8 PostgreSQL-7.3.3 gcc 2.95.3 compiled with readline and fuzzystrmatch without error. After doing an ANALYSE command, any equality comparison between two string fields from different table failed and make the server restart. The PostgreSQL server complains about a shared memory error and panics. The database was exported from 7.2.1 using a custom format dump, and restored into 7.3.3. It is always reproducible anytime after the database was restored using: mydb=# analyse; mydb=# select * from tableA join tableB using (pkeyIntID) where tableA.varcharField = tableB.varcharField; -> connection dropped, server restarted -> #! prompt appears There is an unique btree index on (pkeyIntID, varcharField) defined on both tables. I tried to rebuild database indexes, vacuum full, vacuum analyse, in many combinaisons, it changed nothing, the bug was still there. So, anybody has a clue? If you need more information, just let me know. Thanks a lot, Patrick _________________________________________________________________ Add photos to your e-mail with MSN 8. Get 2 months FREE*. http://join.msn.com/?page=features/featuredemail
"Patrick Vachon" <kaleidelf@hotmail.com> writes:
> Solaris 8
> PostgreSQL-7.3.3
> After doing an ANALYSE command, any equality comparison between two string
> fields from different table failed and make the server restart.
If you are running in a non-C locale, you might see whether the problem
goes away after re-initdb'ing in C locale. I have some recollection of
strcoll() having bugs in various versions of Solaris. It'd be worth
trying to get a stack trace out of the core dump file to verify where
it's crashing, too.
regards, tom lane
I initdb'ed with --no-locale option and that fixed the problem. Thanks a lot Tom! Should this issue be documented in src/doc/FAQ_Solaris? IMHO, I think it should, since it may affect various versions of Solaris. Regards, Patrick >From: Tom Lane <tgl@sss.pgh.pa.us> >To: "Patrick Vachon" <kaleidelf@hotmail.com> >CC: pgsql-ports@postgresql.org >Subject: Re: [PORTS] ANALYSE on Solaris 8 bug Date: Mon, 30 Jun 2003 >12:54:40 -0400 > >"Patrick Vachon" <kaleidelf@hotmail.com> writes: > > Solaris 8 > > PostgreSQL-7.3.3 > > > After doing an ANALYSE command, any equality comparison between two >string > > fields from different table failed and make the server restart. > >If you are running in a non-C locale, you might see whether the problem >goes away after re-initdb'ing in C locale. I have some recollection of >strcoll() having bugs in various versions of Solaris. It'd be worth >trying to get a stack trace out of the core dump file to verify where >it's crashing, too. > > regards, tom lane > >---------------------------(end of broadcast)--------------------------- >TIP 8: explain analyze is your friend _________________________________________________________________ Tired of spam? Get advanced junk mail protection with MSN 8. http://join.msn.com/?page=features/junkmail
"Patrick Vachon" <kaleidelf@hotmail.com> writes:
> Should this issue be documented in src/doc/FAQ_Solaris?
> IMHO, I think it should, since it may affect various versions of Solaris.
If we knew enough about it to document what combinations work and what
don't, I'd have put something in before now. I'm not going to put in
something as vague as "non-C locales may not work on Solaris", because
I don't think the problem is nearly as sweeping as that.
If you care to troll our archives and whatever info is available from
Sun, and put together a useful entry for the FAQ, that'd be great.
regards, tom lane
Here what I found: http://www.science.uva.nl/pub/solaris/solaris2/Q5.2.html Since 7.3.3 now use Solaris qsort, that may explain the problem I had. Hope this help, Patrick >From: Tom Lane <tgl@sss.pgh.pa.us> >To: "Patrick Vachon" <kaleidelf@hotmail.com> >CC: pgsql-ports@postgresql.org >Subject: Re: [PORTS] ANALYSE on Solaris 8 bug Date: Mon, 30 Jun 2003 >16:28:09 -0400 > >"Patrick Vachon" <kaleidelf@hotmail.com> writes: > > Should this issue be documented in src/doc/FAQ_Solaris? > > IMHO, I think it should, since it may affect various versions of >Solaris. > >If we knew enough about it to document what combinations work and what >don't, I'd have put something in before now. I'm not going to put in >something as vague as "non-C locales may not work on Solaris", because >I don't think the problem is nearly as sweeping as that. > >If you care to troll our archives and whatever info is available from >Sun, and put together a useful entry for the FAQ, that'd be great. > > regards, tom lane > >---------------------------(end of broadcast)--------------------------- >TIP 4: Don't 'kill -9' the postmaster _________________________________________________________________ Tired of spam? Get advanced junk mail protection with MSN 8. http://join.msn.com/?page=features/junkmail
"Patrick Vachon" <kaleidelf@hotmail.com> writes: > Here what I found: > http://www.science.uva.nl/pub/solaris/solaris2/Q5.2.html Hmm ... that's suspicious, but it's talking about Solaris 2.5. Surely Sun woulda fixed that problem by Solaris 8? regards, tom lane