#asahi-alt · 2024-05-05 09:01
leiomakes it better though#asahi-alt · 2024-05-05 09:01
chadmedprobably doesnt help performance that i have energy efficient workqueues on by default ;)#asahi-alt · 2024-05-05 09:01
leioI know, I'm running it; it doesn't solve it here#asahi-alt · 2024-05-05 09:00
chadmedjanne made dcp's rtkit threads explicitly ask for high priority workqueues#asahi-alt · 2024-05-05 09:00
leioI'm comparing with how mutter gets realtime prio#asahi-alt · 2024-05-05 09:00
sam_I'll think about it here (not looked at the bug in detail yet, just saw it in inbox)#asahi-alt · 2024-05-05 09:00
leiois cmake adding CAP_SYS_NICE when it opts to depend on libcaps?#asahi-alt · 2024-05-05 09:00
chadmedthe rtkit involved here is the apple one, not the poettering one#asahi-alt · 2024-05-05 09:00
sam_especially whre it might lead to people rpeorting stuff upstream#asahi-alt · 2024-05-05 08:59
chadmedbut there will probably be some people grumbling about building with caps#asahi-alt · 2024-05-05 08:59
sam_for KDE stuff we try to be opinionated about avoiding broken cases#asahi-alt · 2024-05-05 08:59
leiohow does that work there anyways vs rtkit?#asahi-alt · 2024-05-05 08:59
chadmedi dont really care enough to voice dissent if thats what asturm wants to do#asahi-alt · 2024-05-05 08:59
chadmedyeah #asahi-alt · 2024-05-05 08:58
leio(go with that in main tree)#asahi-alt · 2024-05-05 08:58
leioah, you don't think we should go with that plan, misread part of it.#asahi-alt · 2024-05-05 08:57
leiowith just kwin[caps] it'll break if that plan comes to fruition in main tree.#asahi-alt · 2024-05-05 08:57
leiocaps(+) means that it will assume caps is there when kwin loses IUSE=caps, which seems to be the current plan.#asahi-alt · 2024-05-05 08:57
chadmedi just did kwin[caps]. i dont think it should be unconditional, but +caps in IUSE is a good solution#asahi-alt · 2024-05-05 08:56
leiohe looks to be looking feedback from sam_ in relation to that#asahi-alt · 2024-05-05 08:56
leioif asturm will make it just force enabled always, you might want to make the dep to be kwin[caps(+)] so it doesn't break when it happens (but I haven't looked what you did)#asahi-alt · 2024-05-05 08:55
chadmedso no user intervention should be required if they are installing plasma properly#asahi-alt · 2024-05-05 08:55
chadmedjannau: i made the asahi-meta USE flag "kde" since that gets selected by the desktop/plasma profile anyway#asahi-alt · 2024-05-05 08:54
sam_firefox requires care because of the stupid rust thing#asahi-alt · 2024-05-05 08:54
chadmedyeah i tested all good :) #asahi-alt · 2024-05-05 08:51
leioit's not like I know for sure that `if use foo && (use bar || use baz)` will work in ebuild ;p#asahi-alt · 2024-05-05 08:43
LarstiQlooking forward to it :)#asahi-alt · 2024-05-05 08:42
chadmedLarstiQ: *everything* about $thing is absolutely awful. im writing a blog post about just how abjectly awful it is in fact since im allowed to now#asahi-alt · 2024-05-05 08:40
LarstiQchadmed: oh that sounds awful || behaviour of $thing#asahi-alt · 2024-05-05 08:38
chadmedim adding llvm:18 to LLVM_COMPAT btw #asahi-alt · 2024-05-05 08:30
chadmed$thing at $work is ambiguous about ||'s behaviour. sometimes its XOR and sometimes it's not and now i have gaslit myself into second guessing every || operator#asahi-alt · 2024-05-05 08:30
chadmedyeah dw dont listen to me#asahi-alt · 2024-05-05 08:28
chadmedlike its not XOR in bash#asahi-alt · 2024-05-05 08:27
chadmedhang on really stupid question i think ive overthought this but (use opencl || use video_cards_asahi) is still true when BOTH are true right#asahi-alt · 2024-05-05 08:27
chadmedboth work totally fine with llvm:18 #asahi-alt · 2024-05-05 08:27
chadmedsame with ff#asahi-alt · 2024-05-05 08:27
leiowants to get rid of llvm:17 :D#asahi-alt · 2024-05-05 08:26
leiodoesn't like how conservative in-tree mesa is about LLVM_COMPAT#asahi-alt · 2024-05-05 08:26
chadmedyeah already got that one#asahi-alt · 2024-05-05 08:25
leiootherwise there was also the dev-libs/libclc[spirv(-)] BDEPEND shared with intel, which you probably got already too#asahi-alt · 2024-05-05 08:24
leioguess we don't need rust at all for mesa yet?#asahi-alt · 2024-05-05 08:23
chadmedyeah thats the plan. the only shared dep here is spirv-llvm-translator so im just going to refactor the if use llvm && use opencl bit and that fixes it#asahi-alt · 2024-05-05 08:23
leiook, but I guess if we need spirv-llvm-translator stuff outside opencl, you'll need to repeat those and the necessary deps (presumably not all that opencl wants) and then a bunch if `if use llvm && (use opencl || use video_cards_asahi)` or something?#asahi-alt · 2024-05-05 08:22
chadmedif we explicitly depend on spirv-llvm-translator outside of opencl then we should probably express that a bit nicer than just duplicating all the llvm deps#asahi-alt · 2024-05-05 08:21
chadmedwell i want to make the ebuild have as few modifications from upstream as possible. now that llvm deps are handled (mostly) sanely with LLVM_SLOT etc all i have done is add video_cards_asahi ( llvm ) to REQUIRED_USE#asahi-alt · 2024-05-05 08:21
leioI don't think our spirv-llvm-translator stuff is for opencl?#asahi-alt · 2024-05-05 08:20
leioin the old ebuild you separately made asahi do llvm stuff too; are you moving towards controlling it via USE=llvm and making it REQUIRED_USE instead?#asahi-alt · 2024-05-05 08:20
chadmedyeah that got it#asahi-alt · 2024-05-05 08:18
chadmedlooks like opencl is masked (???) so its not picking up PKG_CONFIG_PATH #asahi-alt · 2024-05-05 08:18
chadmedif use llvm && opencl