19:45 letterrip: hi all, is there a simple test suit to determine which calls are causing the desktop to freeze?
19:45 letterrip: suite
19:46 letterrip: I'm getting a freeze where the mouse pointer still works, but the desktop and apps are unresponsive
19:47 letterrip: which i can only recover from either restarting lightdm or rebooting
19:48 letterrip: i can't switch to virtual terminals (ctrl alt f1 etc.) and can only use a terminal via ssh
19:48 letterrip: i'm using the xfce (xubuntu) desktop
19:49 letterrip: and have disabled compositing
19:49 letterrip: which makes the lockups less frequent
19:50 letterrip: i'm using disco so that I have the most recent kernel (5.0) and the oibaf build that contains the latest mesa drivers
19:52 RSpliet: letterrip: it's unlike that it's directly relatable to "a call", due to the nature of command buffers. "The desktop to freeze while the mousepointer still works" is sadly a very generic symptom that can be triggered by many problems.
19:52 RSpliet: Some have been hunted down over the years, some are still somewhat a mystery
19:52 letterrip: RSpliet, i figured it was fairly generic
19:53 RSpliet: If you look through the logs, the first error is likely to be "CTXSW_TIMEOUT"
19:53 letterrip: was hoping there might be a fuzzer or such for discovering the causes of grpahics bugs
19:53 letterrip: dmesg or which logs?
19:53 RSpliet: dmesg yeah
19:53 RSpliet: If I'm right, it;s sadly just as uninformative. It simply means the card ended up in a state it shouldn't be, and we don't know how to recover from it
19:55 letterrip: nope that isn't what i have - I get 'failed to idle channel'
19:55 letterrip: well actually that isn't from my typical problem just now
19:55 letterrip: i just tried to load blender
19:55 letterrip: which much more reliably causes crashes :)
19:55 letterrip: but likely seperate issue
19:56 letterrip: RSpliet, is there a detailed debugging that can be done?
19:57 letterrip: I'm probably willing to put forth the effort if there is detailed instructions somewhere
19:57 RSpliet: letterrip: I'm sorry, I'm unaware of specific procedures you could follow unless there is a very deterministic path to failure
19:57 RSpliet: And I suspect this is a case of "once/twice a day, when the stars align, it randomly occurs"
19:59 letterrip: this looks promissing - https://www.phoronix.com/scan.php?page=news_item&px=GraphicsFuzz-Demo
20:00 letterrip: https://github.com/google/graphicsfuzz
20:02 RSpliet: letterrip: problem is that we're probably quite aware this can generate many many crashes and other undesirable behaviours. But there's no engineering time to analyse the data and really figure out what the fuzzer triggered.
20:02 RSpliet: (when I say we, I mean they. I haven't had time for nouveau in the past 2 years
20:03 letterrip: RSpliet, right... well it appears to also provide a method for simplifying
20:04 letterrip: anywho I'll go ahead and play with it and see if I get anything useful
20:08 karolherbst: well we suck if one application causes the GPU to hang