Re: Intermittent meson failures on msys2

[Thomas Huth <thuth@redhat.com>]

On 08/07/2022 15.13, Daniel P. Berrangé wrote:
> On Fri, Jul 08, 2022 at 04:41:48PM +0400, Marc-André Lureau wrote:
>> Hi
>>
>> On Mon, Jun 27, 2022 at 6:41 AM Richard Henderson <
>> richard.henderson@linaro.org> wrote:
>>
>>> Hi guys,
>>>
>>> There's an occasional failure on msys2, where meson fails to capture the
>>> output of a build
>>> script.  E.g.
>>>
>>> https://gitlab.com/qemu-project/qemu/-/jobs/2642051161
>>>
>>> FAILED: ui/input-keymap-qcode-to-linux.c.inc
>>> "C:/GitLab-Runner/builds/qemu-project/qemu/msys64/mingw64/bin/python3.exe"
>>> "C:/GitLab-Runner/builds/qemu-project/qemu/meson/meson.py" "--internal"
>>> "exe" "--capture"
>>> "ui/input-keymap-qcode-to-linux.c.inc" "--"
>>> "C:/GitLab-Runner/builds/qemu-project/qemu/msys64/mingw64/bin/python3.exe"
>>> "../ui/keycodemapdb/tools/keymap-gen" "code-map" "--lang" "glib2"
>>> "--varname"
>>> "qemu_input_map_qcode_to_linux" "../ui/keycodemapdb/data/keymaps.csv"
>>> "qcode" "linux"
>>> [301/1665] Generating input-keymap-qcode-to-qnum.c.inc with a custom
>>> command (wrapped by
>>> meson to capture output)
>>> ninja: build stopped: subcommand failed.
>>>
>>>
>>> https://gitlab.com/qemu-project/qemu/-/jobs/2625836697
>>>
>>> FAILED: ui/shader/texture-blit-frag.h
>>> "C:/GitLab-Runner/builds/qemu-project/qemu/msys64/mingw64/bin/python3.exe"
>>> "C:/GitLab-Runner/builds/qemu-project/qemu/meson/meson.py" "--internal"
>>> "exe" "--capture"
>>> "ui/shader/texture-blit-frag.h" "--" "perl"
>>> "C:/GitLab-Runner/builds/qemu-project/qemu/scripts/shaderinclude.pl"
>>> "../ui/shader/texture-blit.frag"
>>> [313/1663] Generating texture-blit-vert.h with a custom command (wrapped
>>> by meson to
>>> capture output)
>>> ninja: build stopped: subcommand failed.
>>>
>>>
>>> Could you have a look please?
>>>
>>>
>>>
>> Ah, we don't have artifacts for msys2 builds it seems, that would perhaps
>> help. It would make sense to at least take meson-logs/*.txt. I'll work on a
>> patch.
>>
>> My guess is that CI randomly fails with "too many opened files", as I have
>> seen that regularly on various projects with Windows runners. And here,
>> it's probably reaching limits when running python/perl scripts
>> simultaneously... I don't see an easy way to solve that if that's the issue.
> 
> There shouldn't be very much parallelism even taking place, because
> 
> https://docs.gitlab.com/ee/ci/runners/saas/windows_saas_runner.html
> 
> says  "Windows runners execute your CI/CD jobs on n1-standard-2
>         instances with 2 vCPUs and 7.5 GB RAM. "
> 
> unless ninja is setting a parellism much higher than nCPUs ?

We're compiling with "make -j2" there, see .gitlab-ci.d/windows.yml ... so I 
don't think that it's about too many things going on in parallel. 
Additionally, the problem hasn't been there a couple of weeks ago, so it's 
either something new that we merged recently, or the Windows containers or 
MSYS2 environment has been changed recently?

  Thomas