RE:PostgreSLQ v10.1 and xlC compiler on AIX - Mailing list pgsql-hackers

From REIX, Tony
Subject RE:PostgreSLQ v10.1 and xlC compiler on AIX
Date
Msg-id B37989F2852398498001550C29155BE51790B5C4@FRCRPVV9EX3MSX.ww931.my-it-solutions.net
Whole thread Raw
In response to Re: PostgreSLQ v10.1 and xlC compiler on AIX  (Michael Paquier <michael.paquier@gmail.com>)
List pgsql-hackers
Hi Michael,

Thanks for the information ! I'm not a PostgreSQL expert.

I've found the file:  ./32bit/src/test/regress/regression.diffs

Since I'm rebuilding right now, I do not have just now the file for v10.1 with the issue.
However, I have it for 10.0 and 9.6.6 which show the same issues. I've attached them to this email.

About the suggestion of adding a AIX machine with xlc v13.1 to the buildfarm, I'll see how this could be done.
Also, I'll try to get a more recent version of xlC v13 .


Cordialement,

Tony Reix

Bull - ATOS
IBM Coop Architect & Technical Leader
Office : +33 (0) 4 76 29 72 67
1 rue de Provence - 38432 Échirolles - France
www.atos.net

________________________________________
De : Michael Paquier [michael.paquier@gmail.com]
Envoyé : jeudi 23 novembre 2017 14:08
À : REIX, Tony
Cc : PostgreSQL Hackers; OLIVA, PASCAL
Objet : Re: PostgreSLQ v10.1 and xlC compiler on AIX

On Thu, Nov 23, 2017 at 7:57 PM, REIX, Tony <tony.reix@atos.net> wrote:
> We are porting PostgreSQL v10.1 on AIX (7.2 for now).
> And we have several tests failures, in 32bit and 64bit.
> We are using xlc 13.01.0003.0003 with -O2.
> Tests were 100% OK with version 9.6.2 .
>
> About 32 bit failures within v10.1, we have 3 failures including:
>      create_aggregate         ... FAILED
>      aggregates               ... FAILED
>
> I've found that these 2 failures disappear when building :
>    ./src/backend/parser/gram.c
> without -O2 !
> However, this is a 44,498 lines file and it may take a while for finding the
> root issue.

When running regression tests and those fail, there is a file called
regressions.diff which gets generated in src/test/regress showing the
difference between the results generated and the results expected when
running the SQL queries which are part of the regression tests.
Attaching this file to this thread would help in determining what's
wrong with the regression tests.

> I invite anyone involved in porting/using PostgreSQL 10.1 on AIX to take
> part of a discussion about how to make v10.1 work perfectly on AIX.

Note that xlc 12.1 is tested with AIX machines on the buildfarms, but
there is no coverage for 13.1 visibly. It would be nice if you could
set up a buildfarm machine to catch problems way earlier.
--
Michael

Attachment

pgsql-hackers by date:

Previous
From: Alvaro Herrera
Date:
Subject: Re: PostgreSLQ v10.1 and xlC compiler on AIX
Next
From: Peter Eisentraut
Date:
Subject: documentation is now XML