#1142192 vkd3d: FTBFS on ppc64el: multiple failures in tests/d3d12

Package:
src:vkd3d
Source:
src:vkd3d
Submitter:
Simon McVittie
Date:
2026-07-17 12:57:01 UTC
Severity:
normal
Tags:
#1142192#5
Date:
2026-07-16 11:18:26 UTC
From:
To:
Thanks for merging the new upstream release of vkd3d. Unfortunately, the
new release fails several of its automated tests in the "tests/d3d12"
test-case on ppc64el:

https://buildd.debian.org/status/fetch.php?pkg=vkd3d&arch=ppc64el&ver=2.0%2Bds-1&stamp=1784146331&raw=0

If I'm reading the log correctly, the test failures are:

I'm not familiar enough with this codebase and architecture to know
whether this is normal for ppc64el, and I don't have any ppc64el
hardware myself.

Before 2.0, in practice the test suite probably wasn't run at all,
because Mesa GL/Vulkan drivers weren't installed - so this test failure
might not actually be a regression. If that's the case, then perhaps
these three test-cases should be ignored/marked as "todo" on ppc64el for
now?

    smcv
#1142192#10
Date:
2026-07-16 12:19:07 UTC
From:
To:
This is reproducible in a schroot environment on the ppc64el porterbox
"platti", if that helps.

     smcv
#1142192#17
Date:
2026-07-16 19:01:56 UTC
From:
To:
I would assume so, unless someone can confirm the test was run before.

Probably, yes. Interesting that it fails with llvmpipe on a specific architecture, but for floating point corner cases it makes sense.
#1142192#22
Date:
2026-07-16 19:01:56 UTC
From:
To:
I would assume so, unless someone can confirm the test was run before.

Probably, yes. Interesting that it fails with llvmpipe on a specific architecture, but for floating point corner cases it makes sense.
#1142192#27
Date:
2026-07-17 12:55:11 UTC
From:
To:
Control: tags -1 + patch

I confirm that the test was *not* run before. In the build log
https://buildd.debian.org/status/fetch.php?pkg=vkd3d&arch=ppc64el&ver=1.2-15%2Bb3&stamp=1768648399&raw=0
the tests appear to have passed, but actually if you look at the
individual .log files (which unfortunately are not available in older
buildd logs) you'll see that most (all?) tests were skipped because
there was no working Vulkan driver. This can be reproduced by rebuilding
1.2-15 in a sid chroot on the ppc64el porterbox, platti.

A build-dependency on Vulkan and GL drivers was added in
https://salsa.debian.org/wine-team/vkd3d/-/commit/d2dcaf91440111271a285b119d2a05ba8409a8b6
with the result that most of the tests are now run (under
llvmpipe/lavapipe).

https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/2139 or the
attached patch successfully works around this (confirmed on platti).

     smcv