01:05 <tlwoerner> i would, but i've yet to reproduce reliably, seems like every time i run the full glmark2-es2 test suite at --fullscreen it fails somewhere but in slightly different ways

01:06 <tlwoerner> if i run the [terain] test by itself, even at --fullscreen, it does politely say the test is unsupported and quits

01:07 <tlwoerner> but when [terain] is run as part of the entire suite, the board seems to lock up (i'll have to confirm by plugging in ethernet). in any case the console is unresponsive

01:07 <tlwoerner> (again, at --fullscreen) when run normally it seems to behave okay and it behaves okay when run --off-screen as well

02:10 <anarsoul> check dmesg?

02:14 <anarsoul> tlwoerner: anyway, it's very unlikely that it's related to terrain

09:32 <enunes> tlwoerner: yeah all my tests just ran the default set including [terrain] which just checks that it is not supported and skips, it shouldn't do any harm

09:47 <enunes> but nice to see that someone got motivated to do some more checking based on the talk

09:48 <enunes> other applications than glmark2 again would be nice, it is not as easy of course but well, thats why it would be a contribution

10:53 <tlwoerner> a couple years ago i stumbled across the same behaviour in etnaviv and it turned out it was a bug. there's a difference between running tests one at a time versus altogether. not enough something-or-others were being freed between tests

10:54 <tlwoerner> enunes: are you running glmark2 --fullscreen on a rock64?

10:56 <tlwoerner> i've never been able to get the full suite run --fullscreen and start the test after [terrain]

10:56 <tlwoerner> https://paste.ee/p/KyB6h

10:57 <tlwoerner> it runs [terrain], reports that it's not supported, then the system halts

10:58 <tlwoerner> if i knew how to enable more debugging i would give that a try

11:00 <enunes> tlwoerner: I'm not running with --fullscreen, I can try that (a bit later)

11:00 <tlwoerner> running without --fullscreen works for me :-)

11:01 <enunes> I don't have a rock64, but hopefully the hardware shouldnt matter all that much, and if it is a platform specific thing then it might be that it is not something too relevant for lima unfortunately

11:01 <enunes> so hopefully if it is a problem it does reproduce on other platforms

11:05 <tlwoerner> it would be great to know, either way :-)

11:06 <tlwoerner> what platform do you have/test? are there other tests you suggest/recommend?

14:41 <rellla> enunes: i somehow fixed the clip_control issue. deqp and piglit results are online. now i get the depth_clamp test from skip to fail...

14:42 <rellla> have to check either how to not expose GL_ARB_depth_clamp or fix it ;)

14:43 <rellla> http://lima.imkreisrum.de/piglit/a2c30c1..a2c30c1-main/results.json.bz2 is the raw result of mesa main. would be cool if you can do a diff with your results

14:51 <rellla> MR is updated, found some new bits in render state...

16:25 <anarsoul> tlwoerner: it worked for me on rock64 couple months ago

16:26 <anarsoul> tlwoerner: do you have a heatsink on the cpu?

16:26 <tlwoerner> anarsoul: yes, a tiny, passive one

16:26 <anarsoul> same here

16:28 <anarsoul> are you using wayland, X11 or run it on plain drm?

16:29 <anarsoul> I presume X11 or wayland since drm won't accept --fullscreen option

16:30 <anarsoul> tlwoerner: ^^

16:31 <tlwoerner> X11, es2 "flavour" (for glmark2)

16:31 <tlwoerner> oops, x11-glesv2 flavour

16:33 <anarsoul> tlwoerner: how much memory your rock64 has?

16:34 <anarsoul> 1gb may be running out of memory in fullscreen in 1080p

16:34 <anarsoul> but again - try ssh-ing into the box and grab dmesg

16:35 <tlwoerner> according to u-boot: DRAM: 4 GiB

16:35 <tlwoerner> okay