00:01 dj-death: have people noticed issues with bindgen?
00:02 dj-death: like :
00:02 dj-death: panicked at 'Unable to generate bindings: ClangDiagnostic("error: unknown argument: '-mtls-dialect=gnu2'\n")
00:02 dj-death: it's bit weird because this flag shouldn't be there if I look at the meson code
00:02 dj-death: so not sure what inserts it
01:53 Liver_K: Ah I see, https://en.wikipedia.org/wiki/X.Org_Server#Device_Dependent_X_(DDX)
02:26 Liver_K: karolherbst: Can you point me to any resource that would help me to configure nouveaue to use the DDX module?
03:17 fdobridge_: <a​irlied> okay found the xfb bug at least
03:18 fdobridge_: <a​irlied> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26558
03:42 fdobridge_: <a​irlied> Pass: 8944, Fail: 11, Warn: 1, Skip: 90, Timeout: 3, Flake: 1, Duration: 15:06, Remaining: 0
06:45 fdobridge_: <a​irlied> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26559 one more xfb bug
06:47 Liver_K: I still need help using the modesetting DDX karolherbst described
06:50 airlied: it should be automatic, just don't have an xorg.conf
06:51 Liver_K: I do not
06:52 airlied: then X should pick modesetting I think, or deinstall the xorg-x11-drv-nouveau or whatever the distro calls it
06:55 Liver_K: ... Back to square one then, see my original problem
06:58 Liver_K: I don't have an xorg.conf anywhere or any kind of package that includes "drv"
06:58 Liver_K: In the name*
07:05 airlied: can you pastebin xorg log file?
07:34 Liver_K: airlied: Sure, https://0x0.st/H3Hh.0.log
07:40 Liver_K: So looks like it is already loading the modesetting_drv.so module, whatever that is
08:01 airlied: indeed looks like it should, so probably not whatever karolherbst thought it might e
08:01 airlied: be
08:05 Liver_K: So apart from what might be causing tearing and window border trailing, I'm wondering now what these *_drv.so modules are, and why the nouveau one isn't being loaded
08:15 Liver_K: Also I'm not sure this is the right log file lol, it says it is initializing a Logitech M525 pointing device, which I have not plugged into this machine for months, if ever
08:20 Liver_K: airlied: I got that one straight from /var/log/Xorg.0.log, is it supposed to be somewhere other than there?
08:30 fdobridge_: <a​irlied> Might be one in .local somewhere
10:47 karolherbst: "[ 23.491] (==) Log file: "/var/log/Xorg.0.log", Time: Wed Aug 9 16:04:15 2023"
13:43 Liver_K: airlied: Ah there is one in .local/share/xorg that makes more sense, here it is: https://0x0.st/H3KQ.0.log
14:32 karolherbst: Liver_K: yeah, looks alright
14:33 karolherbst: not sure if Option "TearFree" "true" is set by default though...
14:34 karolherbst: yeah.. it's not
14:34 karolherbst: mhh
15:38 fdobridge_: <g​fxstrand> I wonder if NVIDIA's DNZ bit also flushes inf and NaN
15:54 fdobridge_: <k​arolherbst🐧🦀> good question
15:54 fdobridge_: <k​arolherbst🐧🦀> docs say it's only limited to products with 0
15:57 fdobridge_: <g​fxstrand> kk
16:29 fdobridge_: <g​fxstrand> This should make D3D9 go a tiny bit faster. 😁
16:29 fdobridge_: <g​fxstrand> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26569
16:59 fdobridge_: <m​arysaka> nice!
17:57 fdobridge_: <k​arolherbst🐧🦀> I finally got the GV100 :3
18:00 fdobridge_: <k​arolherbst🐧🦀> @gfxstrand which branch should I check with?
18:00 fdobridge_: <k​arolherbst🐧🦀> and what should I check? 😄
18:13 fdobridge_: <b​utterflies> do you have pmu firmware? 😄
18:15 fdobridge_: <g​fxstrand> Do a full CTS run on main
18:23 fdobridge_: <k​arolherbst🐧🦀> 1.0? or 1.1?
18:23 fdobridge_: <k​arolherbst🐧🦀> or everything?
18:25 fdobridge_: <g​fxstrand> It should test 1.1
18:25 fdobridge_: <g​fxstrand> But use deqp-runner because there will be fails.
18:25 fdobridge_: <g​fxstrand> Most of those fails will be a wrong if statement somewhere
19:22 fdobridge_: <a​irlied> @gfxstrand marge for 26558 okay now?
20:03 Lyude: I see why ben was hoping that we wouldn't have to handle specific errors from GSP calls lol
20:03 Lyude: this is kind of a pain to handle in C
20:19 fdobridge_: <a​irlied> I think that is all the low hanging GL CTS fixes, or at least the ones where I don't need zmike to consult
20:26 fdobridge_: <g​fxstrand> Oh, and `NVK_USE_NAK=all` if you aren't using a my draft Volta MR.
21:51 fdobridge_: <a​irlied> @gfxstrand https://paste.centos.org/view/raw/b2c507b3 right or shouldn't be necessary, I think it's a zink bug but the test passes and doesn't crash with just that change
21:59 fdobridge_: <d​adschoorse> zink should check shaderStorageImageMultisample, which nvk doesn't support yet?
22:04 fdobridge_: <g​fxstrand> What's crashing on that? That looks very wrong. We don't support storageImageMultisample (yet)
22:07 fdobridge_: <a​irlied> KHR-GL45.gl_spirv.spirv_validation_capabilities_test test, and it does spit out warnings about missing stuff. but this is the only place it dies, but I suspect I might need to fix something in zink
22:13 Lyude: things aren't working quite yet but I'm definitely on the right track I think
22:13 Lyude: I'm seeing us manage to recover from aux failures now
22:14 Lyude: but I'm missing something regarding where we toggle the LCD_VCC line through gsp
22:15 fdobridge_: <g​fxstrand> I made the mistake of looking at an NV shader dump. They read the descriptor, decode it, and swizzle manually. 😭
22:16 airlied: Lyude: so you know how to toggle it, just not sure where to do it from?
22:17 Lyude: airlied: something like that. I think there might be one more delay or something I'm missing. So far I've found at least two timeouts (one that GSP provides from the aux methods, one from the link training methods) - so I think there's another timeout I need around main_link_ctrl (the gsp callback that I assume is toggling LCD_VCC)
22:19 Lyude: there's a bunch of different ways nvidia seems to handle that. Some panels want main_link_ctrl toggled on/off, some don't, some want it after DPCD power on, some want it before… etc
22:21 Lyude: i'm definitely getting there though :), I just need to do more tracing
22:23 Liver_K: karolherbst: How would I set that option? Just in an xorg.conf?