03:01 imirkin: AndrewR: it's not to be tonight. i'm going to just send a revert tomorrow and work it out another day.
03:02 AndrewR: imirkin, ok (I hope testing / ci will not break hard)
03:02 imirkin: i'll just have to disable those jobs. not a big deal - they were fairly new anyways
14:46 yuken: Huh. Does nouveau not do >1080p60 on Pascal?
14:46 imirkin: should work
14:46 imirkin: what seems to be the problem?
14:49 yuken: My screen goes dark (still visible) and it... looks similar to an interlaced display, in terms of a sort of blur, imirkin.
14:49 yuken: Installing the proprietary drivers now to make sure it isn't my card.
14:55 yuken: Yep, appears to be an issue with nouveau. Gentoo 5.10.10, GTX 1050, HDMI, to an MSI AG32C.
14:58 RSpliet: yuken: when running nouveau, can you install edid-decode, then do... eh..
14:58 RSpliet: cat /sys/class/drm/card0/card0-HDMI-<somethingsomething>/edid | edid-decode
14:59 RSpliet: and post the result of that on a paste website, share the link here
14:59 yuken: RSpliet, not right now, but I'll gladly do that tonight.
15:00 RSpliet: hmm, that may not tell us much about what nouveau does with those modes though...
15:19 imirkin: yuken: you didn't mention what you were trying to do
15:20 imirkin: i.e. what mode are you trying to set?
15:20 imirkin: and what kernel?
15:20 imirkin: RSpliet: fyi: https://people.freedesktop.org/~imirkin/edid-decode/ -- can take output from xrandr
15:20 yuken: 5.10.10, 1920x1080 120, 144, or 165, progressive.
15:20 imirkin: yuken: i've heard of issues with 144+ refresh rates, but no clue why
15:20 yuken: Happens at 120 too.
15:20 imirkin: none of the developers have such a screen, so tricky to debug
15:21 imirkin: others have gotten 120hz to work
15:21 RSpliet: imirkin: I was looking for a method that could also work with wayland :-P
15:22 imirkin: RSpliet: just an fyi. you can also upload a file into it. or you can run stuff locally too, just some people may not have the edid-decode tool
15:23 RSpliet: ah oh you made an on-line version of that. Neat-o!
15:23 imirkin: yeah, was play with wasm
15:24 imirkin: RSpliet: also https://people.freedesktop.org/~imirkin/nvbios/
15:24 RSpliet: Yeah I remembered that one :-) Useful when I'm too lazy to git pull
16:19 imirkin: karolherbst: i'm going to nuke the runner on nouveau for now -- i have to revert the fence change for now
16:19 karolherbst: okay
16:19 imirkin: we can have another go at it later on
16:20 karolherbst: imirkin: what's actually the problem without the change? runtime stuck forever or what's up?
16:21 imirkin: after-the-change, the problem is channel failure + recovery fail
16:21 imirkin: before-the-change, the problem is that the shim will cause the code to hang on screen destroy
16:22 imirkin: (since it waits for a fence, which is obviously not gonna happen with the shim)
16:22 imirkin: originally i tried to find a way to make it detect that the shim is being used
16:22 imirkin: but i couldn't really work out a reasonable way of doing it
16:32 imirkin: stupid smart meters
16:45 RSpliet: stupid sexy Flanders
16:47 imirkin: hehe
16:47 imirkin: feels like wearing nothing at all!
17:03 karolherbst: imirkin: we could start a thread in the shim to just signal fences
17:03 karolherbst: or just signal them on submission
17:03 imirkin: how do we know it's a fence?
17:03 imirkin: it's just some bo where the code is expecting an int to increase
17:03 karolherbst: right
17:04 karolherbst: mhh..
17:04 karolherbst: yeah.. would be annoying to parse the push buffer :/
17:23 imirkin: karolherbst: in the meanwhile: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8867
17:33 karolherbst: imirkin: why not keep the function but just change it?
17:33 imirkin: coz i'm reverting the commit
17:33 karolherbst: mhh
17:42 imirkin: i think it's a reasonable argument, i'll have a look
20:17 AndrewR: imirkin, also, stupid idea, but can't you use some env. variable to tell driver when use old way of doing things 9on real hw) vs running emulation (with drm-shim).?
20:18 imirkin: AndrewR: that's super-inelegant...
20:18 imirkin: i want something which Just Works (tm)
20:18 AndrewR: imirkin, then good luck (sorry I can't help there)
20:19 imirkin: yeah no worries
20:19 imirkin: priority is to (a) get mesa going again
20:19 imirkin: then (b) repro locally
20:19 imirkin: and then (c) figure out a thing
20:22 AndrewR: imirkin, but if you never reach (b) ?
20:22 imirkin: then someone who cares about drm-shim in CI will figure it out
20:22 AndrewR: imirkin, sounds like 2 or 3 ppl at most .. ;/
20:22 imirkin: welcome to nouveau.
20:23 AndrewR: imirkin, yeah, I tend to gravitate into corners
20:25 AndrewR: I mean there is this effect when (understanable) ppl mostly use more stable/performant ("better") driver, but this effect tend to suck testing etc away from any even temporary disadvantaged component ..so, I try to resist this?
20:26 imirkin: the tests are new and low-impact in the first place
20:26 imirkin: i added drm-shim because (a) eric asked for it and (b) seemed like fun
20:26 imirkin: i even learned something about the nouveau ABI in the process
20:27 imirkin: (or perhaps ... re-discovered)
20:27 AndrewR: :}