Feedback-ID: i131946ab:Fastmail
Message-ID: <78d1cfa6-0065-865d-584b-cde6d8c18aff@enterprisedb.com>
Date: Tue, 25 Apr 2023 12:27:52 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:102.0)
 Gecko/20100101 Thunderbird/102.10.0
Subject: Re: ssl tests aren't concurrency safe due to get_free_port()
Content-Language: en-US
To: Andrew Dunstan <andrew@dunslane.net>, Andres Freund <andres@anarazel.de>
Cc: pgsql-hackers@postgresql.org, Tom Lane <tgl@sss.pgh.pa.us>
References: <20221002164931.d57hlutrcz4d2zi7@awork3.anarazel.de>
 <7579f627-aad4-7a04-1a98-033cc6b3cca5@dunslane.net>
 <52266bf8-bf38-fdfa-9a77-1568f82b63b3@dunslane.net>
 <0891f64a-d30f-a1b7-9d0c-e447684cd8cd@dunslane.net>
 <b8f5d215-cd90-9ea3-88d8-03a2c864ebb1@dunslane.net>
 <20221105183643.v2nxsac3xlb3eazi@awork3.anarazel.de>
 <0867ca05-ed3f-09ca-cd41-d7087e542362@dunslane.net>
 <e47aa1f7-039e-77a3-89fb-ee77dbb20a7c@dunslane.net>
 <20221116015105.g6qxu2ugpj6agt26@awork3.anarazel.de>
 <58de442f-a4fb-3980-85c8-a024ff26c54b@dunslane.net>
 <20221119201615.uy4rvspdq62oqikn@awork3.anarazel.de>
 <41ba523e-c867-2c04-3c34-d283a6ba9f7b@dunslane.net>
From: Peter Eisentraut <peter.eisentraut@enterprisedb.com>
In-Reply-To: <41ba523e-c867-2c04-3c34-d283a6ba9f7b@dunslane.net>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Archived-At: <https://www.postgresql.org/message-id/78d1cfa6-0065-865d-584b-cde6d8c18aff%40enterprisedb.com>
Precedence: bulk

On 20.11.22 16:10, Andrew Dunstan wrote:
> 
> On 2022-11-19 Sa 15:16, Andres Freund wrote:
>> Hi,
>>
>> On 2022-11-19 10:56:33 -0500, Andrew Dunstan wrote:
>>>> Perhaps we should just export a directory in configure instead of this
>>>> guessing game?
>>> I think the obvious candidate would be to export top_builddir from
>>> src/Makefile.global. That would remove the need to infer it from
>>> TESTDATADIR.
>> I think that'd be good. I'd perhaps rename it in the process so it's
>> exported uppercase, but whatever...
>>
> 
> OK, pushed with a little more tweaking. I didn't upcase top_builddir
> because the existing prove_installcheck recipes already export it and I
> wanted to stay consistent with those.
> 
> If it works ok I will backpatch in couple of days.

These patches have affected pgxs-using extensions that have their own 
TAP tests.

The portlock directory is created at

     my $build_dir = $ENV{top_builddir}
       || $PostgreSQL::Test::Utils::tmp_check ;
     $portdir ||= "$build_dir/portlock";

but for a pgxs user, top_builddir points into the installation tree, 
specifically at $prefix/lib/pgxs/.

So when running "make installcheck" for an extension, we either won't 
have write access to that directory, or if we do, then it's still not 
good to write into the installation tree during a test suite.

A possible fix is

diff --git a/src/Makefile.global.in b/src/Makefile.global.in
index 5dacc4d838..c493d1a60c 100644
--- a/src/Makefile.global.in
+++ b/src/Makefile.global.in
@@ -464,7 +464,7 @@ rm -rf '$(CURDIR)'/tmp_check && \
  $(MKDIR_P) '$(CURDIR)'/tmp_check && \
  cd $(srcdir) && \
     TESTDIR='$(CURDIR)' PATH="$(bindir):$(CURDIR):$$PATH" \
-   PGPORT='6$(DEF_PGPORT)' top_builddir='$(top_builddir)' \
+   PGPORT='6$(DEF_PGPORT)' top_builddir='$(CURDIR)' \
     PG_REGRESS='$(top_builddir)/src/test/regress/pg_regress' \
     $(PROVE) $(PG_PROVE_FLAGS) $(PROVE_FLAGS) $(if 
$(PROVE_TESTS),$(PROVE_TESTS),t/*.pl)
  endef