[Postgres-xl-developers] Postgres-xl on Power: AIX and Linux/Power
REIX, Tony
tony.reix at atos.net
Thu Dec 14 10:05:33 PST 2017
Hi Tomas,
1) Linux/PPC64 BE
---------------------------
I have Flex v2.6.0 installed on my Linux/PPC64 BE, and that is an issue for the configure file, which does not handle this case. I've patched it.
flex-2.6.0-3.fc25.ppc64
I got the master from https://github.com/Postgres-XL/Postgres-XL ,
and I have:
...
============== creating temporary instance ==============
============== initializing database system ==============
sh: /root/rpmbuild/BUILD/Postgres-XL-master/contrib/btree_gin/log/initgtm.log: No such file or directory
pg_regress: initgtm failed
2) AIX
I have:
gcc --version
gcc (GCC) 8.0.0 20171205 (experimental)
However, the issue deals with the configure*/Makefile* files not handling correctly AIX.
The issue I described in previous email is due to "ld on AIX not working like ld works on Linux I think. I had to add -Wl,-G to say "ld" to accept missing definitions of symbols.
Moreover, LDFLAGS="........." ./configure does not work as expected. A GNUmakefile file identical to GNUmakefile.in is generated, but no Makefile files was regenerated.
I guess I have to do some aclocal/autoconf/automake, however that broke the ./configure .
I've found some other issues on AIX, that i'll have to fix later.
The most important issue is to understand why ./configure does not work with postgres-XL on AIX. With posgresSQL, we use the official source code and we simply do (in short): ./configure ... ; gmake ; ... and that works fine.
Cordialement,
Tony Reix
ATOS / Bull SAS
ATOS Expert
IBM Coop Architect & Technical Leader
Office : +33 (0) 4 76 29 72 67
1 rue de Provence - 38432 Échirolles - France
www.atos.net
________________________________________
De : Tomas Vondra [tomas.vondra at 2ndquadrant.com]
Envoyé : jeudi 14 décembre 2017 18:33
À : REIX, Tony; postgres-xl-developers at lists.postgres-xl.org
Cc : EMPEREUR-MOT, SYLVIE; OLIVA, PASCAL
Objet : Re: [Postgres-xl-developers] Postgres-xl on Power: AIX and Linux/Power
Hi,
Can you please run the tests on the latest master sources, and share the
regression.diffs from those machines?
On 12/14/2017 04:51 PM, REIX, Tony wrote:
> Hi,
>
> Some more details about v10alpha2:
> -*FC25 / Power LE (GCC 7.1.0):*
> - 20 failures (inc. 2 _random_ failures):
> brin collate combocid _create_index_ date groupingsets
> inherit limit privileges rowtypes rules sanity_check select_parallel
> sequence stats subselect tsrf updatable_views _window_ with
>
> - *FC25 / Power BE (GCC 6.3.1):*
> - a *hang* once
> - up to 37 failures:
> brin cluster collate combocid constraints copy2 create_index
> groupingsets identity indirect_toast inherit insert insert_conflict join
> limit plpgsql privileges rangefuncs returning rowtypes rules
> sanity_check select_parallel sequence stats stats_ext subselect truncate
> tsrf union updatable_views with xc_distkey xc_sequence xl_functions
> xl_limitations xl_user_defined_functions
>
> So. It seems that there are 19 failures appearing only within BigEndian.
> I'm interested in getting these failures fixed some day, since AIX also
> is BE.
Thanks for doing those tests. I see those tests were executed on alpha2,
so certain number of failures are expected. But I wonder why the BE gets
almost twice as many of them.
If you observe a "hang" again, can you check stacktraces of the stuck
processes? I guess you observed a process waiting for incomplete GTM
response - that was a bug in alpha2, but it should be fixed in head.
> On AIX, I have 0 failure with PostgreSQL v9.6.6 & v10.1 and XLC.
Yeah, there's still some fallout from the PostgreSQL 10 merge. I agree
we need to clean that up soon.
regards
--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
More information about the Postgres-xl-developers
mailing list