[06:18] <alkisg> I've been testing 16 bpp depth with the modesetting driver, and it has issues everywhere that I tested it: intel => crash, vboxvideo (vbox) and virtio (kvm) => corrupted double image, cirrus (kvm) => corrupted small image etc
[06:18] <alkisg> But there's something I don't understand. Ubuntu 16.04.1 and 16.04.2 have the same modesetting driver, right?
[06:19] <alkisg> So, why e.g.  `kvm -vga cirrus -ctrl-grab -m 1024 -cdrom ubuntu-mate-16.04.1-desktop-i386.iso` works, while 16.04.2 has issues?
[06:37] <alkisg> Hey, it's Xorg 1.18.3 vs 1.18.4, let me test if that's to blame....
[06:46] <alkisg> So, I've booted with ubuntu-mate-16.04.2-desktop-i386.iso which does have the 16 bpp issue, then I downgraded from xserver-xorg-hwe-16.04 (1.18.4) to xserver-xorg (1.18.3), and it still has the issue. What does that prove, that the issue is in the cirrus driver and not in the modesetting driver?!
[06:46] <alkisg> (while 16.04.1 with xserver-xorg (1.18.3) does not have the issue)
[06:47] <alkisg> (unless installing a different xorg requires a reboot too, not just a xinit...)
[06:51] <alkisg> Sorry my bad, I didn't notice that the xserver-xorg-core version was unrelated to the xserver-xorg-* stack. By installing xserver-xorg-core=1.18.3, the problem disappeared
[06:52] <alkisg> So I believe that the regression happened somewhere between 1.18.3 and 1.18.4. Testing more...
[07:17] <alkisg> I filed this one: https://bugs.freedesktop.org/show_bug.cgi?id=100295
[07:17] <ubottu> Freedesktop bug 100295 in Driver/modesetting "Regression in 16bpp between 1.18.3 and 1.18.4" [Normal,New]
[07:18] <alkisg> What real hardware can I test, that would use the modesetting driver? Except for intel, which crashed and I've already reported that?
[07:23] <tjaalton> whatever you have
[07:31] <alkisg> tjaalton: I can specify "Driver modeset" in xorg.conf in e.g. a radeon laptop that I have here?
[07:31] <tjaalton> yes
[07:31] <alkisg> ty, testing...
[08:57] <alkisg> I updated the bug report with the results of the real hardware test, which are similar.
[08:58]  * alkisg checks if he can find any recent commits that could have caused this...
[09:00] <tjaalton> there aren't too many between 1.18.3..1.18.4
[09:00] <tjaalton> bisect
[09:01] <tjaalton> or just try reverting a commit at a time that touches modesetting
[09:01] <tjaalton> 62 commits in 1.18.4
[09:02] <tjaalton> 11 on modesetting
[09:07] <alkisg> tjaalton: I'm a little lost between apt source and upstream git, is it possible to git/bzr clone and bisect something from launchpad instead? E.g. is there a bzr branch of xserver-xorg-hwe-16.04 in launchpad?
[09:07] <tjaalton> no
[09:08] <alkisg> I'm seeing this: https://cgit.freedesktop.org/xorg/xserver/log/hw/xfree86/drivers/modesetting, and I can't find out which commit maps to the 1.18.3 release
[09:09] <tjaalton> git clone upstream, git log xorg-server-1.18.3..xorg-server-1.18.4
[09:09] <tjaalton> then start reverting the commits that touch modesetting, starting from the latest
[09:09] <alkisg> ty, trying...
[09:09] <tjaalton> next git format-patch -1 <sha>, drop the patch in debian/patches and add to series. build a package
[09:10] <tjaalton> it's probably f091528457cc62f modesetting: Implement 32->24 bpp conversion in shadow update
[09:30] <alkisg> Yup, that was the one
[09:30] <alkisg> Thank you Timo, commenting on the report...
[09:42] <alkisg> tjaalton: eh, it's also the one that's causing the broadwell crash on 16bpp :D
[09:43] <tjaalton> figured
=== JanC_ is now known as JanC