On 17. Apr 2022, at 5.43, Noah Meyerhans <no...@debian.org> wrote:
>
> On Wed, Apr 13, 2022 at 09:33:58PM +0200, Christian Göttsche wrote:
>> It looks like the ppc64el build was retried and succeeded.
>>
>> Maybe the timeouts should be increased for slow build systems, similar
>> what upstream did for running under valgrind with commit 2d12409a
>> ("lib-smtp: Adjust test timeouts based on valgrind runtime
>> presence")[1].
>
> Build logs are available at (1) and don't look like a timeout. It was
> rebuilt on a different machine (2), though, which could be relevant. I
> wonder if there's a different configuration. Every dovecot build on
> mipsel-manda-04 has failed since 2021-09-04, across a few different
> versions. We have seen one failure on mipsel-osuosl-04 that was
> successful on retry, though, so it could be a red herring.
>
> I haven't had a chance to look into this more deeply.
>
> 1.
> https://buildd.debian.org/status/fetch.php?pkg=dovecot&arch=mips64el&ver=1%3A2.3.18%2Bdfsg1-1%2Bb1&stamp=1647208211&raw=0
> 2. https://buildd.debian.org/status/logs.php?pkg=dovecot&arch=mips64el
The test-program-client-net is listen()ing on 127.0.0.1 in
test_program_setup(). Then in test_program_refused() it tries to connect to ::1
-> 127.0.0.3 -> 127.0.0.2 -> 127.0.0.1 and expects that only the last one
succeeds and the others fail with "connection refused". But for some reason
more than the last of those connects are succeeding. So bind()ing to the
listener socket somehow isn't working as expected.
