07:28 imirkin: kisak: confirmed that steam loads fine with my changes on a nv42. going to do some more testing though
08:10 imirkin: hrmph. what happened to auto-fan control on nv4x?
08:33 mupuf: imirkin: it's supposed to work
08:34 imirkin: iirc there were some bugs filed about it stopping to work
08:34 imirkin: you even looked into it, no?
08:39 imirkin: mupuf: drm/nouveau: Don't disable polling in fallback mode
08:39 imirkin: looks like still unmerged
08:39 mupuf: shit, really?
08:39 mupuf: ...
08:40 imirkin: or at least not in my kernel, which is 4.19 :)
08:40 mupuf: I thouht I was done with it, just waiting on skeggsb to merge
08:40 imirkin: yeah
08:40 imirkin: he missed it? or something.
08:41 imirkin: maybe send it to airlied direct?
08:44 mupuf: I'll check it out a bit later. I am trying hard to get the new membership website for xorg ready in time
08:46 imirkin: sure
15:15 orbea: imirkin: any chance you might have a guess why RetroArch is now failing with KMS + nouveau + wayland on fedora? Someone reported this to me with this log: https://pastebin.com/raw/XywMFjA0 It was fixed poorly for me with amd + xorg here which appears to be the first bad commit for them https://github.com/libretro/RetroArch/pull/7708/commits/5898f3e5d22b930a1050a59b61e98ecd07dd6977 and then a better fix
15:15 orbea: shortly after here https://github.com/libretro/RetroArch/pull/7729/commits/868465ad01dba03b5f86f185dfe357c2eb94dc6a The original issue was here https://github.com/libretro/RetroArch/issues/7119
15:20 orbea: they have a GK107
16:24 glennk: orbea, see footnote 8 in https://www.khronos.org/registry/EGL/specs/eglspec.1.5.pdf
16:24 glennk: section 3.4
16:25 orbea: glennk: thanks, looking
17:41 imirkin: is that the footnote that says "p.s. nouveau doesn't work"? :)
17:43 imirkin: orbea: iirc i looked at that commit and it seemed buggy
17:44 orbea: the on by maister?
17:44 orbea: *one
17:45 orbea: i'm sure if the issue was pointed out to maister it could be solved, but I dont have any direct means of contacting him...i tried here and he might see it there... https://github.com/libretro/RetroArch/pull/7729#issuecomment-450574422
17:45 imirkin: the second one is weird
17:45 imirkin: but it does seem like it could work
17:46 imirkin: orbea: is there any info on the failure?
17:46 orbea: all I have are the logs here https://pastebin.com/raw/XywMFjA0
17:48 orbea: gouchi: is the one who is having the issue if there is anything more detailed to get.
17:49 imirkin: orbea: [INFO] Environ SET_PIXEL_FORMAT: RGB565.
17:49 imirkin: however the code very explicitly looks for 8888 configs
17:49 orbea: ah...that makes sense why it would blow up
17:50 imirkin: i don't know retroarch, but perhaps that's a line of investigation
17:50 imirkin: i.e. i don't know what SET_PIXEL_FORMAT does, but one could reasonably imagine it has something to do with pixel formats :)
17:52 orbea: i would imagine before the first buggy commit (Idk why things get merged and not tested until it blows up...) it was just finding the correct format
17:53 imirkin: yeah, there's not a lot of options for RGB565
17:53 imirkin: just that and BGR565
17:53 imirkin: since it probably pegs green at >= 6 pixels
17:53 imirkin: but the GBM format is different for that
17:54 imirkin: this is why i didn't want to contribute a patch in the first place - lots of unknowns :)
17:54 orbea: yea, i know the feeling...there is a lot going on a way too many supported platforms...
17:54 glennk: there's byte swapped rgb565 too...
17:55 imirkin: that's just mean
17:58 imirkin: i've gotta say ... i'm impressed how unbuggy the chromium experience is on a NV42
17:58 imirkin: google maps gets a bit flickery
17:58 imirkin: and there's some obvious misrenders, but the idea is there
17:58 imirkin: but most regular pages are totally fine
17:59 imirkin: aha. Rasterization: Software only. Hardware acceleration disabled =]
17:59 glennk: was about to mention that :-)
17:59 imirkin: still nice.
18:00 glennk: its likely the best option on that era hardware
18:00 imirkin: for some reason this GPU *has* to be the only one plugged into PCIe
18:00 imirkin: normally i keep these types of things secondary, but the box hangs on boot if i don't with this one - FX 3450
18:01 glennk: it uses a agp bridge chip doesn't it?
18:01 imirkin: nah, native PCIe
18:01 imirkin: NV42
18:02 imirkin: GPU core: +1.22 V (min = +4294967.28 V, max = +1.37 V)
18:02 imirkin: that min seems a bit high
18:02 glennk: ah, nv40 was the one then with the bridge
18:02 imirkin: and oddly reminiscent of MAX_UINT
18:02 glennk: then they had the same chip in the other direction for a while for agp variants
18:02 imirkin: glennk: and NV4A which is also AGP-native, although there's a version of it with PCI (but no PCIe)
18:03 imirkin: (aka NV44A)
18:04 imirkin: but yeah, there were also random nv4x's on PCIe -> AGP bridges. i haven't seen many of them though.
18:04 imirkin: (i think there might even have been a handful of G80+ boards like that)
18:05 imirkin: i try to stay away from AGP - generally go the other way. i even have a PCIe FX5200
18:05 imirkin: (but not the PCIe GF4)
19:39 joepublic: AGP sure was exciting at the time
20:05 glennk: burn with firewire
20:52 imirkin: urgh. this is sad. xonotic eventually runs into some various data corruption :(
20:53 imirkin: [on nv42]
21:00 imirkin: and of course the corruption doesn't reproduce in a trace