[libvirt] build fails randomly while generating docs

[Biswa96]

Description / Steps to reproduce the issue

libvirt fails to build docs randomly. Just enable the docs option or remove -Ddocs=disabled option from meson command, if any. This issue was first reported here #13426

Expected behavior

Docs should be build without any issue.

Actual behavior

The build fails with the errors like following:

[503/812] Generating docs/drvnodedev.html with a custom command (wrapped by meson to capture output)
FAILED: docs/drvnodedev.html
"F:/msys64/ucrt64/bin/meson" "--internal" "exe" "--capture" "docs/drvnodedev.html" "--" "F:\msys64\ucrt64\bin/xsltproc.EXE" "--stringparam" "pagesrc" "docs/drvnodedev.rst" "--stringparam" "builddir" "F:/my-MINGW-packages/mingw-w64-libvirt/src/build-UCRT64" "--stringparam" "timestamp" "Tue Mar  7 16:00:29 2023 UTC" "--stringparam" "href_base" "" "--nonet" "../libvirt-9.1.0/docs/site.xsl" "docs/drvnodedev.html.p/drvnodedev.html.in"
[505/812] Generating 'docs/drvqemu.html.p/drvqemu.html.in' (wrapped by meson to capture output).
ninja: build stopped: subcommand failed.
==> ERROR: A failure occurred in build().
    Aborting...

Also the build fails with different file in each run. Sometimes, the issue can be workaround by running meson compile or ninja twice.

Verification

• I have verified that my MSYS2 is up-to-date before submitting the report (see https://www.msys2.org/docs/updating/)

Windows Version

MINGW64_NT-10.0-19045

MINGW environments affected

• MINGW64
• MINGW32
• UCRT64
• CLANG64
• CLANG32
• CLANGARM64

Are you willing to submit a PR?

No response

[Biswa96]

This issue was brought to attention by @1480c1

@eli-schwartz helped with some meson related issues before. @elmarco also maintains and contributes to virtualization related repositories. They may provide some insights.

[eli-schwartz]

The failing command is:

"F:/msys64/ucrt64/bin/meson" "--internal" "exe" <more arguments>

It fails without emitting any stdout or stderr.

So really, this is just a duplicate of #11864 and related to the fact that python itself sometimes crashes for unknown reasons, non-reproducibly. But it was successfully worked around in CI with 6ff1a1f.

[Biswa96]

That workaround doesn't fix this issue with libvirt. I can reproduce this issue in my PC and in CI.

[eli-schwartz]

Don't know, don't care. :D

The only thing I know is that this is the same issue because I've seen it and dealt with it before, as Meson is often hit by it. Fixing an issue in CPython isn't something I can help you with. All I can do is verify that yes, it's a duplicate bug report and you should redirect your attention to the canonical tracking ticket.

The workaround used to work, but it is after all a workaround, workarounds don't fix anything, so it stopping to work isn't especially surprising.

The other remediation strategy I know of is to just... restart the build, again and again, until it succeeds. It's a flaky issue, if you babysit it long enough it will eventually work. I think I once restarted the Meson CI 10 times until I finally got a passing run.

Obviously that is not a solution either.

[Biswa96]

Understood. Thank you.

Duplicate of #11864

[Biswa96]

I workaround the issue with meson compiler || meson compiler -j1 command.

[Biswa96]

The workaround was removed in this abe8ecf commit and no longer required after python 3.11 update.