00:25karolherbst: Lyude: I am wondering.. maybe we need a call with intel like we do with nvidia, were we can essentially bring random stuff up... or do we have something like that already?
15:03uis: Any news?
15:05karolherbst: uis: only ideas yet
15:05karolherbst: I mean, I can give you a hacky workaround to make it work
15:05karolherbst: but that's not something we should put upstream
16:03PaulePanter: Hi. https://paste.debian.net/1118402/
16:04PaulePanter: Users get that when using RStudio 1.2.1335. Any idea why?
16:04PaulePanter: [ 9290.139130] nouveau 0000:01:00.0: rstudio: Unknown handle 0x0000003b
16:04PaulePanter: [ 9290.139133] nouveau 0000:01:00.0: rstudio: validate_init
16:04PaulePanter: [ 9290.139133] nouveau 0000:01:00.0: rstudio: validate: -2
16:07PaulePanter: On another system with identical configuration:
16:07PaulePanter: [175084.869400] nouveau 0000:01:00.0: fifo: PBDMA0: 00040000  ch 6 [003fa78000 rstudio] subc 0 mthd 0000 data 00000000
16:07PaulePanter: I guess RStudio uses a lot of OpenGL(?) and that is problematic?
16:19PaulePanter: I found https://bugs.freedesktop.org/show_bug.cgi?id=103721 looking quite similar.
16:23karolherbst: PaulePanter: ohh, sounds like multithreading corruption maybe
16:23karolherbst: PaulePanter: do you know if there is a way to disable multithreaded OpenGL in RStudio?
16:24PaulePanter: I’ll take a look.
16:24karolherbst: or maybe it's something else.. but that's my first bet
16:24PaulePanter: Btw: Mesa 19.2.2 is used.
16:25PaulePanter: karolherbst: In one trace there is a crash:
16:25PaulePanter: rstudio: segfault at 980 ip 00007f87e686bd06 sp 00007fff621725b0 error 6 in libdrm_nouveau.so.2.0.0[7f87e6869000+4000]
16:27karolherbst: ahh yeah
16:27karolherbst: then it's most likely the multithreaded issue
16:28PaulePanter: karolherbst: There is *OS Integration*, *Rendering Engine*.
16:29PaulePanter: Options: Auto-detect, Desktop OpenGL, OpenGL for Embedded Systems, Software
16:30PaulePanter: The strange thing is, the PC pool has identical systems. And the problem during a course/class happened only on four or five.
16:30PaulePanter: But I guess multithread issues depend on load too.
16:31karolherbst: yeah.. and it's quite random
16:31karolherbst: if software isn't too slow, you could use that instead
16:32PaulePanter: Is documentation or man-power missing to fix this?
16:32karolherbst: rather, it's not easy fixable
16:32PaulePanter: or hardware
16:32PaulePanter: karolherbst: Is the bug/problem described somewhere in detail?
16:33karolherbst: well, concurrency issues, meaning multiple threads accessing the same data at the same time
16:33karolherbst: it's more of an architectural issue
16:33karolherbst: PaulePanter: mind running with RSTUDIO_CHROMIUM_ARGUMENTS="--disable-gpu" ?
16:33karolherbst: seems that should disable using the GPU when nouveau is in use
16:34karolherbst: or alternativly if you have a "production" system and have to have it work relibly, maybe you should use nvidia by default instead
16:34karolherbst: anyway, we want to fix this issue, it's just not something you spend a week on and it's fixed
16:34karolherbst: it's more like months
16:35karolherbst: PaulePanter: maybe that's related? https://support.rstudio.com/hc/en-us/articles/360017886674-Troubleshooting-RStudio-Rendering-Errors
16:35karolherbst: "Use GPU blacklist (recommended)"
16:35karolherbst: and chromium blacklists nouveau for that very reason as well
16:38karolherbst: anyway.. I have some code fixing it to 99% and I think some of those fixes could actually land before fixing it entirely
16:38karolherbst: as the same needs to be done when fixing it
16:40PaulePanter: karolherbst: Wow, thank you for the details, and working on the fix.
16:40PaulePanter: I’ll can test more tomorrow.
16:41PaulePanter: *GPU blacklist (recommended)* is checked by default.
16:49karolherbst: yeah.. then I guess just the --disable-gpu thing should make it work relible
16:49karolherbst: and GPU acceleration doesn't sound like that important, at least even in chromium it often doesn't make any difference