mbox series

[0/5] sefltest: disable qemu graphics

Message ID 20251017122649.493701-1-mikko.rapeli@linaro.org
Headers show
Series sefltest: disable qemu graphics | expand

Message

Mikko Rapeli Oct. 17, 2025, 12:26 p.m. UTC
RFC: did not have time to properly test this, but sending this out
in case this helps with current release issues

Autobuilders see failures like
https://autobuilder.yoctoproject.org/valkyrie/#/builders/23/builds/2714
and
https://autobuilder.yoctoproject.org/valkyrie/#/builders/67/builds/2630
where selftest are booting qemu which hangs, possibly due to graphics
issues.

Disable graphics from selftests which do not need it. Other selftests
already use runqemu with nographic.

Mikko Rapeli (5):
  selftest devtool.py: disable qemu graphics
  selftest imagefeatures.py: disable qemu graphics
  selftest overlayfs.py: disable qemu graphics
  selftest package.py: disable qemu graphics
  selftest runtime_test.py: disable qemu graphics

 meta/lib/oeqa/selftest/cases/devtool.py       |  2 +-
 meta/lib/oeqa/selftest/cases/imagefeatures.py |  4 ++--
 meta/lib/oeqa/selftest/cases/overlayfs.py     | 12 ++++++------
 meta/lib/oeqa/selftest/cases/package.py       |  4 ++--
 meta/lib/oeqa/selftest/cases/runtime_test.py  | 12 ++++++------
 5 files changed, 17 insertions(+), 17 deletions(-)

Comments

Alexander Kanavin Oct. 17, 2025, 6:27 p.m. UTC | #1
Wait. There is no actual evidence that it is graphics that is causing
the problems, is it? The changes are probably okay, but I would not
want to have this baseless implication in commit messages.

Alex

On Fri, 17 Oct 2025 at 14:27, Mikko Rapeli via lists.openembedded.org
<mikko.rapeli=linaro.org@lists.openembedded.org> wrote:
>
> RFC: did not have time to properly test this, but sending this out
> in case this helps with current release issues
>
> Autobuilders see failures like
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/23/builds/2714
> and
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/67/builds/2630
> where selftest are booting qemu which hangs, possibly due to graphics
> issues.
>
> Disable graphics from selftests which do not need it. Other selftests
> already use runqemu with nographic.
>
> Mikko Rapeli (5):
>   selftest devtool.py: disable qemu graphics
>   selftest imagefeatures.py: disable qemu graphics
>   selftest overlayfs.py: disable qemu graphics
>   selftest package.py: disable qemu graphics
>   selftest runtime_test.py: disable qemu graphics
>
>  meta/lib/oeqa/selftest/cases/devtool.py       |  2 +-
>  meta/lib/oeqa/selftest/cases/imagefeatures.py |  4 ++--
>  meta/lib/oeqa/selftest/cases/overlayfs.py     | 12 ++++++------
>  meta/lib/oeqa/selftest/cases/package.py       |  4 ++--
>  meta/lib/oeqa/selftest/cases/runtime_test.py  | 12 ++++++------
>  5 files changed, 17 insertions(+), 17 deletions(-)
>
> --
> 2.34.1
>
>
> -=-=-=-=-=-=-=-=-=-=-=-
> Links: You receive all messages sent to this group.
> View/Reply Online (#225012): https://lists.openembedded.org/g/openembedded-core/message/225012
> Mute This Topic: https://lists.openembedded.org/mt/115806216/1686489
> Group Owner: openembedded-core+owner@lists.openembedded.org
> Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [alex.kanavin@gmail.com]
> -=-=-=-=-=-=-=-=-=-=-=-
>
Ross Burton Oct. 20, 2025, 8:49 p.m. UTC | #2
On 17 Oct 2025, at 19:27, Alexander Kanavin <alex.kanavin@gmail.com> wrote:
> 
> Wait. There is no actual evidence that it is graphics that is causing
> the problems, is it? The changes are probably okay, but I would not
> want to have this baseless implication in commit messages.

I sort of agree here.  ‘nographic’ turns off the qemu display output but does not disable the virtualised graphics hardware (unlike “novga” which removes display hardware entirely).

Of course, if it’s known that there is a non-trivial overhead in actually outputting the display (which, on the AB, is to a xvfb I believe) then we can still do this.

Also, there’s an argument for nographics (or even novga) in oe-selftest so that people running it locally don’t get qemu’s popping up.  Which I presume happens? My build machines have been headless for many years and I can’t even remember if oe-selftest exposes DISPLAY to the qemu at all.  If it does not then nographic won’t make a difference as the output isn’t active in the first place.

Cheers,
Ross