#asahi-gpu · 2023-03-20 17:50
marcanlet me fix that real quick..#asahi-gpu · 2023-03-20 17:50
marcan(breaks chromium)#asahi-gpu · 2023-03-20 17:50
marcan*that* was the thing I was forgetting.#asahi-gpu · 2023-03-20 17:50
i509vcbbut in the case of radeonsi per what jannau said, wouldn't radeonsi reporting "AMD" be wrong as well given amdvlk is the true amd driver?#asahi-gpu · 2023-03-20 17:50
marcanurgh I think I forgot to backport the dual source blending disable thing#asahi-gpu · 2023-03-20 17:48
jnOpenGL vendor string: AMD#asahi-gpu · 2023-03-20 17:48
jnclient glx vendor string: Mesa Project and SGI#asahi-gpu · 2023-03-20 17:48
jnserver glx vendor string: SGI#asahi-gpu · 2023-03-20 17:48
jnradeonsi reports:#asahi-gpu · 2023-03-20 17:47
i509vcb(does radeonsi report mesa as the vendor? I am not home right now to check that)#asahi-gpu · 2023-03-20 17:46
i509vcbso wgpu is wrong then#asahi-gpu · 2023-03-20 17:44
jannauthat is afaik intentional. also the "vendor" of the opengl driver is mesa and not apple#asahi-gpu · 2023-03-20 17:41
i509vcb(wgpu has to guess for gles which vendor the device is from by renderer name and vendor string)#asahi-gpu · 2023-03-20 17:41
i509vcbI assume that is a bug? It seems to do some funky stuff to wgpu which filters some tests from lavapipe#asahi-gpu · 2023-03-20 17:40
i509vcbI've just noticed that apparently the gles driver reports `Mesa` as the vendor where as on other hardware it would report something like `Intel`#asahi-gpu · 2023-03-20 17:20
alyssamew#asahi-gpu · 2023-03-20 16:52
i509vcbI assume that is already known#asahi-gpu · 2023-03-20 16:52
i509vcbJust updated to the new driver and that single flicker to purple when starting kwin does appear #asahi-gpu · 2023-03-20 13:59
eric_engestromsorry for the noise :]#asahi-gpu · 2023-03-20 13:58
eric_engestromyeah no you're right, meta-package == empty package with depends#asahi-gpu · 2023-03-20 13:58
eric_engestromah yeah, I might be getting meta-packages and package groups confused#asahi-gpu · 2023-03-20 13:40
marcanbut yes, we use actual packages#asahi-gpu · 2023-03-20 13:40
marcanI think eric_engestrom was thinking of package groups, not metapackages#asahi-gpu · 2023-03-20 13:31
_jannau_our meta-packages are empty packages: https://github.com/AsahiLinux/PKGBUILDs/blob/main/asahi-desktop-meta/PKGBUILD#asahi-gpu · 2023-03-20 13:30
eric_engestromhttps://archlinux.org/news/switch-to-the-base-devel-meta-package-requires-manual-intervention/#asahi-gpu · 2023-03-20 13:30
eric_engestromhttps://archlinux.org/news/base-group-replaced-by-mandatory-base-package-manual-intervention-required/#asahi-gpu · 2023-03-20 13:29
eric_engestromarchlinux recently noticed this and had to create the `base` & `base-devel` packages to replace the meta-packages#asahi-gpu · 2023-03-20 13:28
eric_engestromon that note, fyi with meta packages when you change them it only affects the next time someone runs `pacman -S your-meta-package` (basically new installs only); it might be better to make an empty package so that you can update it and existing installs get it in their next `-Syu`#asahi-gpu · 2023-03-20 13:25
eric_engestrom👍#asahi-gpu · 2023-03-20 12:58
marcan*use#asahi-gpu · 2023-03-20 12:58
marcanonce we bless the GPU driver as good enough to us by default (probably real soon now) it'll become a dependency and upstream can't break us any more anyway#asahi-gpu · 2023-03-20 12:57
marcaneric_engestrom: I mean the idea was that the whole thing is *opt-in* right now, that's why our users have upstream mesa in the first place#asahi-gpu · 2023-03-20 10:43
eric_engestrombut ok for your plan :)#asahi-gpu · 2023-03-20 10:43
eric_engestromshadowing upstream is not nice, but neither is upstream breaking us without any reason to even touch that 🙃#asahi-gpu · 2023-03-20 00:34
marcanassuming we keep the -edge system for more bleeding edge stuff, that would be just named `mesa-asahi` at that point#asahi-gpu · 2023-03-20 00:33
marcanwhat we'll do once this is shipped by default is make it a dependency of our desktop metapackage, at which point everyone will be forced to install it (unless they remove the metapackage, which is not supported)#asahi-gpu · 2023-03-20 00:33
marcaneric_engestrom: shadowing upstream packages is not super nice, it makes more sense to set up alternats that provide the original package name (which ours does)#asahi-gpu · 2023-03-20 00:24
i509vcb(Or wherever this stuff lives)#asahi-gpu · 2023-03-20 00:24
i509vcbIf I wanted to test the gpu driver on the literal edge, where is the kernel branch that I would to build to run mesa/main with the new ABI?#asahi-gpu · 2023-03-19 23:31
jannaueric_engestrom: I suppose the idea was to make it optional and easy to revert to the upstream version in the case we screw up#asahi-gpu · 2023-03-19 18:42
eric_engestrommarcan (lina, alyssa): for the manjaro/alarm snafu with enabling asahi in the can't-work-upstream package, this reminds me that I was surprised by the choice to rename the mesa package in the asahi repo instead of naming it `mesa` so that it would shadow the one from alarm; perhaps this should still be done now to avoid this kind of issues in the future? (although I can't think of something else they could do)#asahi-gpu · 2023-03-19 16:22
alyssamanaged to provoke 0x783#asahi-gpu · 2023-03-19 16:20
alyssaat a loss here, where is this missing memory#asahi-gpu · 2023-03-19 16:13
alyssano it's not that either.#asahi-gpu · 2023-03-19 16:12
alyssano this would be pathological even by AGX standards..#asahi-gpu · 2023-03-19 16:11
alyssaunless.#asahi-gpu · 2023-03-19 16:10
alyssathere'sjust, not enough entropy here#asahi-gpu · 2023-03-19 16:10
alyssabut .. I'm not seeing where the cntents go#asahi-gpu · 2023-03-19 16:10
alyssarelating to the bindless samplers#asahi-gpu · 2023-03-19 16:10
alyssa02345 is clearly a table of 8-byte records of some kind