Re: [PULL 00/35] tcg patch queue

[Philippe Mathieu-Daudé]

On 9/20/21 12:07, Peter Maydell wrote:
> On Thu, 16 Sept 2021 at 16:30, Richard Henderson
> <richard.henderson@linaro.org> wrote:
>>
>> The following changes since commit 57b6f58c1d0df757c9311496c32d502925056894:
>>
>>   Merge remote-tracking branch 'remotes/hreitz/tags/pull-block-2021-09-15' 
>> into staging (2021-09-15 18:55:59 +0100)
>>
>> are available in the Git repository at:
>>
>>   https://gitlab.com/rth7680/qemu.git tags/pull-tcg-20210916
>>
>> for you to fetch changes up to 50febfe212f24a9b91b4224d03f653415fddf8e1:
>>
>>   tcg/mips: Drop special alignment for code_gen_buffer (2021-09-16 09:37:39 
>> -0400)
>>
>> ----------------------------------------------------------------
>> Restrict cpu_has_work to sysemu, and move to AccelOpsClass.
>> Move cpu_signal_handler declaration out of target/.
>> Misc tcg/mips/ cleanups.
>>
> 
> This seems to result in a failure on the s390x all-linux-static
> CI job:
> https://gitlab.com/qemu-project/qemu/-/jobs/1604251543
> due to a core dump running the 'trap' test.

I don't have access to that runner, nor to an Ubuntu based one.

I can't reproduce on a RHEL8.5 host.

I ran git-bisect,
- from 57b6f58c1d0 ("Merge remote-tracking branch
'remotes/hreitz/tags/pull-block-2021-09-15' into staging")
- to 62e76dc7dab ("Merge remote-tracking branch
'remotes/rth-gitlab/tags/pull-tcg-20210916' into staging")
using:

$ time make -j 2 -C build qemu-s390x run-tcg-tests-s390x-linux-user

All steps consistently report:

  RUN     TCG tests for s390x-linux-user
  BUILD   s390x-linux-user guest-tests with cc
  RUN     tests for s390x
  TEST    float_convs on s390x
  TEST    threadcount on s390x
  SKIPPED float_convs check on s390x because no reference
  TEST    sha1 on s390x
  TEST    linux-test on s390x
  TEST    testthread on s390x
  SKIPPED signals on s390x because BROKEN awaiting sigframe clean-ups
and vdso support
  TEST    test-mmap (default) on s390x
  TEST    float_madds on s390x
  SKIPPED float_madds check on s390x because no reference
  TEST    hello-s390x on s390x
  TEST    csst on s390x
  TEST    ipm on s390x
  TEST    exrl-trt on s390x
  TEST    exrl-trtr on s390x
  TEST    pack on s390x
  TEST    mvo on s390x
  TEST    mvc on s390x
  TEST    trap on s390x
  TEST    signals-s390x on s390x
[ RUN      ] Operation exception
[       OK ]
[ RUN      ] Translation exception from stg
[       OK ]
[ RUN      ] Translation exception from mvc
[       OK ]
[ RUN      ] Protection exception from stg
[       OK ]
[ RUN      ] Protection exception from mvc
[       OK ]
[  PASSED  ]
  SKIPPED gdbstub test sha1 on s390x because need working gdb
  SKIPPED gdbstub test qxfer-auxv-read on s390x because need working gdb

However I note another job timeouted on the same runner:
https://gitlab.com/qemu-project/qemu/-/jobs/1603171791

Are we running multiple jobs in parallel in the same runner?

> The 'check-acceptance' job also hits a timeout on the emcraft_sf2
> test:
> https://gitlab.com/qemu-project/qemu/-/jobs/1604251596

Not looked at this one yet.