Re: Inscrutable CI jobs (avocado & Travis s390 check-tcg)

[Peter Maydell]

On Thu, 29 Sept 2022 at 14:16, Thomas Huth <thuth@redhat.com> wrote:
>
> On 29/09/2022 15.01, Stefan Hajnoczi wrote:
> > I still don't know what the failures from the Avocado tests are:
> > https://gitlab.com/qemu-project/qemu/-/jobs/3100466535
> > https://gitlab.com/qemu-project/qemu/-/jobs/3100466546
> > https://gitlab.com/qemu-project/qemu/-/jobs/3100466539
> > https://gitlab.com/qemu-project/qemu/-/jobs/3100466548
> > https://gitlab.com/qemu-project/qemu/-/jobs/3100466537
> >
> > There are lots of logs and I don't know what to look for.
> >
> > Any ideas?
>
> These are likely the ones I reported here:
>
> https://lists.nongnu.org/archive/html/qemu-arm/2022-09/msg00234.html
>
> Should be fixed by Peter's patches here:
>
> 20220923123412.1214041-1-peter.maydell@linaro.org/T/">https://lore.kernel.org/qemu-devel/20220923123412.1214041-1-peter.maydell@linaro.org/T/

Incidentally, the reason that regression slipped through (which
was my fault) is because I also find Avocado failures
inscrutable -- I do run 'make check-avocado' but I tend to
assume that random timeouts and so on are just flaky tests
rather than actual problems.

In particular in this case QEMU printed an error message
and exited, but the Avocado framework completely failed
to notice this or bring it to anybody's attention -- it
just ends up timing out, which makes the problem much much
less obvious.

-- PMM