Bug#1075713: linux: D-I's X fails to start under kvm -vga qxl
Tj <tj.iam.tj@proton.me> writes:
> Source: debian-installer
> Followup-For: Bug #1075713
> X-Debbugs-Cc: tj.iam.tj@proton.me
>
> I've done some further research via debian-installer repo, build logs,
> and inspecting fb-modules-*-amd64-di packages.
Thanks :-)
> Focusing on just 6.8.12-1 and 6.9.7-1 I cannot see any difference in the
> ISO builds. That is, for both:
>
> * linux-image-*-amd64 does include drivers/gpu/drm/qxl/qxl.ko*
> * fb-modules-*-amd64-di.udeb does not
> * kernel-image-*-amd64-di does not
> * d-i Makefile's DRM_MODULES has/does not list/copy qxl.ko
>
> This makes me wonder if the 20240628_0519 daily netinst really did have
> this module but as I cannot find a copy of the ISO I cannot check.
openQA keeps copies for a while (and that is what date-stamps them, as
the filenames are static[1]):
https://openqa.debian.net/tests/277189/asset/iso/20240628_0519-debian-testing-amd64-netinst.iso
> If it did not then something must have changed in the qemu/kvm side.
Looking at this again:
https://openqa.debian.net/tests/278609#investigation
I realise that it's got a "Show more" thing in the
diff_packages_to_last_good which I'd not noticed earlier. :-/
I _had_ thought that it was just a few apt & autoconf tools that had
changed, but actually there's loads of things, including qemu.
Looking at the etckeeper log on the worker, I see that there was an
upgrade from qemu 1:7.2+dfsg-7+deb12u5 to 1:7.2+dfsg-7+deb12u at
2024-06-29 11:12:50, which is just after the last succesful run, so that
does point towards qemu being the problem.
This afternoon I may have time to check if I can demonstrate that this
is actually a regression in qemu, and will reasign the bug if so (I
don't mind at all if someone else does that test for me :-) )
Cheers, Phil.
--
Philip Hands -- https://hands.com/~phil
Reply to: