13:50 zmike: eric_engestrom: maybe you able to repro the crash in https://gitlab.freedesktop.org/mesa/mesa/-/jobs/61845655 on one of your setups? I'm down to ci printf debugging with no leads
14:15 zmike: dj-death: maybe you have kbl around? it happens there too ^
15:17 dj-death: zmike: not right now
15:17 dj-death: zmike: spec@egl_mesa_query_driver@conformance,Crash ?
15:17 zmike: yes
15:18 zmike: or maybe just test on any of your machines with that branch
15:20 dj-death: let me see
15:32 dj-death: zmike: passes fine on sim
15:32 dj-death: zmike: 2 valgrind warnings
15:33 dj-death: zmike: https://paste.mozilla.org/uyByYvxG
15:33 zmike: hmm don't think those are related but should probably be fixed
15:34 dj-death: annoying to repro the CI's environment...
15:38 zmike: ohhh I found a machine where it repros I think!
15:39 zmike: wtf
15:45 zmike: aha!
15:45 zmike: fucking hell
16:00 dj-death: zmike: what is it?
16:00 zmike: driconf parsing bullshit
16:00 zmike: memory management is hard.
16:01 zmike: pop quiz: you dlopen a lib, dlsym a symbol, call the symbol which returns statically allocated string values, dlclose the lib, and then access the strings
16:01 zmike: what happens
16:10 alyssa: you cry
16:12 zmike: I hate pipe-loader so much
16:12 zmike: how long is this gonna take me to delete...
16:20 zmike: mareko: I added a couple commits to the base of the latest refactor MR, please re-confirm your rb
16:26 MrCooper: zmike: strdup'ing the strings isn't feasible?
16:26 karolherbst: zmike: we need to nuke clover to delete the dynamic pipe-loader 🙃
16:26 karolherbst: (or maybe just make it use the static one)
16:27 karolherbst: and I think once the dynamic one is gone, we can probably simplify things a lot
16:49 zmike: MrCooper: no, it has to all be the same allocation
16:50 zmike: karolherbst: yes
16:50 zmike: now hurry up
16:58 zmike: eyy it's all green
17:42 sravn: karolherbst: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19385 :-)
17:43 karolherbst: yeah.... but that would regress r600 🙃
17:43 karolherbst: anyway.. should probably land this one first: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27783
17:44 karolherbst: as with that one, moving clover to the static pipeline should already be simple enough
17:44 sravn: Seems like a good plan
17:45 karolherbst: *pipe_loader
17:45 karolherbst: and landing that one depends on https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25840
17:46 karolherbst: (which can be speed up by giving me hardware.. probably)
17:48 sravn: I wish you the best luck, things are newer easy.
17:48 Ermine: Sorry for dumb question, is clover a some implementation of opencl?
17:49 jenatali: Yes
20:29 Lynne: gfxstrand: still no plans for descriptor_buffer on nvk?
20:32 airlied: from nouveau earlier : 04:47 < gfxstrand[d]> Okay, it's EXT_descriptor_buffer time!