Re: test-blockjob: intermittent CI failures in msys2-64bit job

[Vladimir Sementsov-Ogievskiy]

On 17.03.23 15:35, Thomas Huth wrote:
> On 17/03/2023 11.17, Peter Maydell wrote:
> > On Mon, 6 Mar 2023 at 11:16, Peter Maydell <peter.maydell@linaro.org> wrote:
> > > On Fri, 3 Mar 2023 at 18:36, Peter Maydell <peter.maydell@linaro.org> wrote:
> > > > I've noticed that test-blockjob seems to fail intermittently
> > > > on the msys2-64bit job:
> > > >
> > > > https://gitlab.com/qemu-project/qemu/-/jobs/3872508803
> > > > https://gitlab.com/qemu-project/qemu/-/jobs/3871061024
> > > > https://gitlab.com/qemu-project/qemu/-/jobs/3865312440
> > > >
> > > > Sample output:
> > > > | 53/89 ERROR:../tests/unit/test-blockjob.c:499:test_complete_in_standby:
> > > > assertion failed: (job->status == JOB_STATUS_STANDBY) ERROR
> > > > 53/89 qemu:unit / test-blockjob ERROR 0.08s exit status 3
> >
> > > Here's an intermittent failure from my macos x86 machine:
> > >
> > > 172/621 qemu:unit / test-blockjob
> > >             ERROR           0.26s   killed by signal 6 SIGABRT
> >
> > And an intermittent on the freebsd 13 CI job:
> > https://gitlab.com/qemu-project/qemu/-/jobs/3950667240
> >
> > > > > MALLOC_PERTURB_=197 G_TEST_BUILDDIR=/tmp/cirrus-ci-build/build/tests/unit 
> > > > > G_TEST_SRCDIR=/tmp/cirrus-ci-build/tests/unit 
> > > > > /tmp/cirrus-ci-build/build/tests/unit/test-blockjob --tap -k
> > ▶ 178/650 /blockjob/ids
> >             OK
> > 178/650 qemu:unit / test-blockjob
> >             ERROR           0.31s   killed by signal 6 SIGABRT
> > ――――――――――――――――――――――――――――――――――――― ✀  ―――――――――――――――――――――――――――――――――――――
> > stderr:
> > Assertion failed: (job->status == JOB_STATUS_STANDBY), function
> > test_complete_in_standby, file ../tests/unit/test-blockjob.c, line
> > 499.
> >
> >
> > TAP parsing error: Too few tests run (expected 9, got 1)
> > (test program exited with status code -6)
> > ――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――
> >
> > Anybody in the block team looking at these, or shall we just
> > disable this test entirely ?
>
> I ran into this issue today, too:
>
>   https://gitlab.com/thuth/qemu/-/jobs/3954367101
>
> ... if nobody is interested in fixing this test, I think we should disable it...
>
>   Thomas

I'm looking at this now, and seems that it's broken since 6f592e5aca1a27fe1c1f6 
"job.c: enable job lock/unlock and remove Aiocontext locks", as it stops 
critical section by new aio_context_release() call exactly after bdrv_drain_all_and(), so 
it's not a surprise that job may start at that moment and the following assertion fires.

Emanuele could please look at it?



--
Best regards,
Vladimir