00:51haagch: anyone else encoding with vaapi seeing something broken with mesa git?
00:58haagch: looks a bit like it's doing that interlacing stuff, but I have set VAAPI_DISABLE_INTERLACE=1
10:04haagch: this breaks vaapi encoding https://cgit.freedesktop.org/mesa/mesa/commit/?id=c4061bb5fa1c15a77ded16f1de3a4bfa2df20384
10:06haagch: interlaced = ... probably turns interlacing on even when it's disabled by environment variable
10:07haagch: yep, setting interlaced = false; instead works
11:53vedranm: MrCooper: any further ideas for https://bugs.freedesktop.org/show_bug.cgi?id=99353 ?
13:29MrCooper: vedranm: I'm afraid not
13:40vedranm: MrCooper: assuming it never worked, what would a dev do to debug the issue?
13:40vedranm: (I can still try very old distros, but at the moment it was bought only fglrx worked)
13:48iive: vedranm: when kms starts, that is when amdgpu or radeon.ko are loaded,
13:48iive: do you get a text output from the booting kernel?
13:48iive: that's right before X/Wayland start.
13:49vedranm: iive: yep, text console is perfect since 4.something
13:49vedranm: only X/Wayland don't work
13:50iive: ok, does the noise change?
13:50iive: i mean, when something should be drawn, is it drawn wrong or not at all.
13:50vedranm: iive: slightly, over time, like a crazy disco
13:51vedranm: like translating a pixel or two
13:51vedranm: and then back
13:52iive: video cards can display framebuffer from any point of their video memory, page flipping is done by writing a new address of such buffer.
13:53iive: my idea is that the bug is in writing the wrong address, so completely random data is shown.
13:54iive: e.g. radeon,ko has an option use_pflipirq , see if setting different values for it could help.
13:54iive: also, try to disable compositing for X.
13:55iive: `modinfo radeon` would show you all module options and short description.
14:00MrCooper: from the description it seems unlikely to be a page flipping related problem, but starting Xorg 'naked' or just with a simple non-composited session would be interesting, as would be pictures or videos showing the corruption
14:04iive: it doesn't have other modes than glimor. is it possible to have accelmethod none ?
14:04vedranm: iive: mhm, I can try both
14:05vedranm: MrCooper: photos/videos are doable, sure
14:05vedranm: iive: not sure what current Fedora supports, but will check
14:07MrCooper: only amdgpu supports "AccelMethod" "none", though "Accel" "off" would also be interesting
14:12notaz: might be worth trying modesetting and maybe even fbdev Xorg drivers
14:16vedranm: notaz: IIRC fedora uses modesetting on amdgpu kernel driver, if that's the case, then it makes no difference
14:16vedranm: xorg modesetting driver, that is
14:16MrCooper: modesetting sure, though it's unlikely to behave differently from our drivers; fbdev doesn't seem very interesting, since it would hit essentially the same code as fbcon, which is known working
14:17notaz: well it would give a working system at least
14:17vedranm: notaz: I want a working HTPC, not a working headless server :)
14:18MrCooper: "Accel" "off" is probably better for that
18:15minikdo: Hi, does anyone ever tried Radeon r9 270x on linux?
19:10agd5f: haagch, encode issue should be fixed