00:13 airlied: zmike: okay I fixed an even more niche llvmpipe bug than the last one, glsl-fs-fogscale now passes in 128-wide mode :)
00:13 zmike: difficulty: impossible
00:13 zmike: how bored are you
00:16 airlied: jetlag works in mysterious ways
02:17 airlied: okay this one needed a piglit test to trigger it, but 128 and 256 should now be the same, and I can work on 512
02:18 karolherbst: yeah.. soo util_queue_finish is kinda broken... it creates a barrier with num_threads, and the jobs inserted are waiting on the barrier. However, a different thread calling util_queue_kill_threads (through atexit_handler) can cause some of those inserted jobs (through util_queue_finish) to be skipped, which means less than num_threads threads are
02:18 karolherbst: waiting on the barrier leading to a dead lock...
02:22 karolherbst: 1. do we want to support util_queue_finish and the atexit_handler to be called concurrently during application teardown? 2. any fun ideas on how to fix it? 🙃 I could just get rid of the barrier, because it's not entirely clear what purpose that one is supposed to have in the first place.
02:22 karolherbst: maybe I'll just create an MR tomorrow...
02:40 karolherbst: mhh guess the barrier is needed to ensure each thread in the queue is taking the barrier job.. mhh
07:14 dos1: dcbaker: the mesa-26.3.0-rc1 tag is missing quite a few of commits present in the branchpoint tag. looks like it was tagged from a branch from before marge rebased it. is that intentional?
08:40 pq: airlied, awesome :-D
11:58 zmike: jenatali: good news, I just discovered that drirc is broken on Windows (and apparently has been for a long time)
13:41 stefan11111: Could I get some eyes on https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/44996 and https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/45013
13:49 zmike: nowrep: any other comments on interlaced or are you still thinking? (no rush)
13:51 nowrep: yes, one more. sent now
13:52 zmike: oh whoops
14:20 jenatali: zmike: awesome. I'm mostly offline today but I'll look when I can
14:20 zmike: sgtm
14:20 zmike: I have another wgl patch coming for swapchain sync too I think
14:24 dcbaker: dos1: It is. And I am very confused on what happened
14:28 dcbaker: eric_engestrom: I don't konw what happened, but I ended up with a completely broken 26.3 branch last night when I created the branch
14:28 dcbaker: There's a bunch of stuff missing and I somehow ended up with the 27.0 commits in that branch
14:28 dcbaker: and it's not based on the branchpoint tag
14:29 dcbaker: is this the point that we actually just replace the branch and just make -rc2, or do I basically back merge the rest of the commits?
14:29 dcbaker: Not sure what the worse thing is here
14:33 dcbaker: because both options seem frankly bad
14:33 dcbaker: Or anyone else with opinions?
14:35 tannicbrick: Hi, user question about an anv Gen9 hang: Overwatch 2 (DXVK) on Comet Lake GT2 hangs rcs0 with VF/VS busy and the pixel side idle; a CS stall before every VS change prevents it, narrower stalls don't. Is there a known Gen9 rule about changing VS state while earlier draws are still in flight? https://gitlab.freedesktop.org/mesa/mesa/-/work_items/16525
15:06 MrCooper: dcbaker: replacing the 26.3 branch with a proper one seems like the least pain overall to me
15:12 kisak: (peanut gallery) sounds like a valid reason to take the safeties off and force push a new branch into place.
15:15 dcbaker: MrCooper, kisak: I'm leaning that way myself, but I'd need someone with more power to either delete the branch or unprotect it. Presumably just deleting it is easier
15:17 MrCooper: a release manager should be able to do it themselves as far as I'm concerned
15:17 MrCooper: i.e. you should probably have at least maintainer role in the Mesa project
15:18 kisak: ^ I read that as gitlab admin to work the safeties, not social structure.
15:19 MrCooper: if a maintainer can't do that, what's the point of that role?
15:20 MrCooper: seems like something that's better not rushed on a Friday anyway
15:20 dcbaker: zmike is apparently the only maintainer, some of the owners haven't been around in... a while
15:59 mareko: do you mean Mesa project owners
15:59 dcbaker: yeah
15:59 mareko: I'm one
15:59 dcbaker: I haven't seen Ilia or Nicolai around in years
16:00 dcbaker: which was more of an aside of "should people who have moved on continue to have owner status?"
16:02 mareko: probably not
16:10 mlankhorst: Seems like even for security it would be better not to have too many inactive accounts with high privileges. :)
16:14 karolherbst: dcbaker: maybe we should just remove anybody that wasn't active in the last X months (6 or so?)... people can always be added back
16:14 simon-perretta-img: Is there already a more "permanent" (GitLab work item/issue?) place where that + the developer access rules were being discussed, or was that just an earlier discussion in this channel
16:16 karolherbst: I think somebody mentioned that we have like 270 listed developers and.... some were wondering what to do about it
16:19 simon-perretta-img: Right yes
16:20 zmike: simon-perretta-img: are you writing zonk now ?
16:24 simon-perretta-img: No that's just the name I chose for my sleep paralysis demon
16:25 zmike: haha
16:27 simon-perretta-img: He keeps me up at night by whispering things like "Mesa: error: GL_INVALID_OPERATION in glReadPixels(read buffer)", etc.
16:27 zmike: does sound like something zonk would do
16:53 simon-perretta-img: Opened https://gitlab.freedesktop.org/mesa/mesa/-/work_items/16602 to try and condense the above
16:54 simon-perretta-img: (Developer access, etc., not zonk)
16:55 zmike: not zonk not interested
16:58 simon-perretta-img: Understandable
16:58 ccr: zonkers
16:58 zmike: if this shared vulkan api thing works out and someone is actually insane enough to write a (worse) GL frontend for it, the name should definitely be zonk
17:01 dcbaker: So... we'll have zink and zonk. Who's going to write zunk, and what is it going to be?
17:01 dcbaker:will see myself out now
17:01 zmike: have a good weekend
17:02 CounterPillow: I hope zmike wraps glide API over Vulkan next
17:02 zmike: too busy burning tokens on piglit tests
17:02 zmike: no organic meatsack wants to write them so here I am putting my calculator on it
17:04 dcbaker: Honestly... AI is pretty good at writing tests
17:04 zmike: the best is when you tell it to wargame tests vs driver
17:08 ccr: "GREETINGS, PROFESSOR ZMIKE. A STRANGE GAME. THE ONLY WINNING MOVE IS NOT TO PLAY."
17:08 zmike: I'm not playing
17:08 zmike: I leave that to the thinking sand
17:08 CounterPillow: what do they pay you for then
17:13 dcbaker: zmike has been promoted to a Project Manager
17:13 zmike: oh no not again
17:14 dcbaker: It's inevatable, AI is coming for all of the programming. We'll all just be PMs in the end
17:14 zmike: I will say
17:14 zmike: being able to orchestrate from the gym is a genuine QoL improvement
17:14 zmike: previously I could only shitpost and try to unbury my inbox, but now I can actually keep working between sets
17:14 zmike: the day when I work entirely from the gym is not far off
17:16 simon-perretta-img: Zunk can be the translation layer for the ancient powervr sgl api, its only purpose will be allowing the world to once again experience the best graphics demo ever created:
17:16 simon-perretta-img: "A seal swimming quickly in polluted water in a sinus pattern was used to demonstrate this card."
17:16 zmike: hahahaa
17:16 zmike: Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
17:17 zmike: okay, so like...serious question now
17:17 zmike: I have https://gitlab.freedesktop.org/mesa/piglit/-/merge_requests/1247
17:17 zmike: is anyone really going to review this many new tests
17:18 zmike: there are no tests for these extensions anywhere
21:42 mareko: I can ask my assistant to review them
21:44 zmike: sounds good 🤝
21:44 zmike: in return, I will have my assistant fix your sqtt
21:44 mareko: I think pepp is the de-factor maintainer of sqtt
21:44 zmike: yeah I posted a MR and you guys can figure it out