#1093520 puma: FTBFS: test failures

Package:
src:puma
Source:
src:puma
Submitter:
Emilio Pozuelo Monfort
Date:
2025-05-28 12:47:02 UTC
Severity:
normal
Tags:
#1093520#5
Date:
2025-01-19 14:35:52 UTC
From:
To:
Some tests are failing for puma at least on amd64 and mips64el:

Errors & Failures:

  1) Failure:
TestIntegrationPumactl#test_refork_cluster [/build/reproducible-path/puma-6.4.3/test/test_integration_pumactl.rb:127]:

Logs on https://buildd.debian.org/status/package.php?p=puma

Cheers,
Emilio

#1093520#12
Date:
2025-04-05 18:32:15 UTC
From:
To:
Hi,

Both bugs are about similar issues (tests failures during build for the
first one, tests failures during autopkgtest for the second one).

Regarding the first one, I could build puma fine in two different
environments, so I think that the failures are random.
Regarding the second one, it was already known that the failures are
random.

I'm thus downgrading the severity of both bugs. Regarding the second
one, I'm not 100% sure: I thought random failures were not
release-critical, but I see that the bug was originally filed as
serious, even if it was already about random failures. So feel free
to upgrade to serious again!

Best,

Lucas

#1093520#17
Date:
2025-04-05 18:53:20 UTC
From:
To:
Hello.

On AWS machines of type m7a.medium and r7a.medium
I'm getting a failure rate of 100%, so it's
not really "random".

Incidentally, those instance types have a single-CPU, but at this
point it's not clear yet if that's the condition which triggers
the build failure or, for example, the fact that they are overall slower
than other instance types, or maybe some other reason.

In either case, we don't have a Build-CPU control field to indicate
that the package "needs" more than 1 CPU to build successfully,
so this is a "secret sauce" that we should try to avoid in a stable
release, as it's against the spirit of the DFSG.

My suggestion is that you try building the package
after booting with GRUB_CMDLINE_LINUX="nr_cpus=1"
and it you can't reproduce the build failures that
way please contact me privately and I will gladly
provide a VM to reproduce.

If you don't want to debug this, that's ok, but then
please disable the flaky tests.

Thanks.

#1093520#22
Date:
2025-04-05 18:56:11 UTC
From:
To:
Oh, I see that Lucas is speaking as the maintainer in this thread.

Could you please build the package on an instance
of type m7a.medium or r7a.medium?

(This time I don't need to offer a VM :-)

Thanks.

#1093520#29
Date:
2025-04-09 08:25:29 UTC
From:
To:
Hi,

Yes (I was going through Ruby team's RC bugs)

I tried with a local KVM virtual machine, with nr_cpus=1, and can
confirm that it fails to build in that case.

Still, it's probably OK to keep it as severity:important in the context
of the trixie release preparation. (or severity:serious +
trixie-ignore). There are many other more severe RC bugs to work on
unfortunately.

Lucas

#1093520#34
Date:
2025-05-28 12:44:28 UTC
From:
To:
retitle 1093520 puma: FTBFS: test failures
thanks

Hello. While it's true that this package seems to fail systematically on single-cpu systems,
the failing test in those cases (single-cpu) is this one:

TestWebConcurrencyAuto#test_web_concurrency_with_concurrent_ruby_available

which is not the error reported by Emilio.

I'm therefore restoring the old title and will report the single-cpu failure separately.

Thanks.