00:00 DavidHeidelberg[m]: zmike: spotted twice recently on the `zink-anv-tgl`: KHR-Single-GL46.enhanced_layouts.ssb_member_align_non_power_of_2,Timeout
00:00 zmike: DavidHeidelberg[m]: flake it
00:00 DavidHeidelberg[m]: s/recently/today/
00:00 DavidHeidelberg[m]: zmike: kk
00:00 zmike: or maybe skip
00:00 zmike: probably skip
00:01 DavidHeidelberg[m]: I'll add it to my MR in queue as a new commit
00:01 zmike: ack
00:05 DavidHeidelberg[m]:started making sandwitch merge requests
00:05 zmike: porkchop sandwiches?!
00:06 DavidHeidelberg[m]:would prefer beef sandwiches, but still sounds great :D
16:30 DavidHeidelberg[m]: enunes: how much of the timeout do you think it would be appropriate to give to Weston? It seems to me from looking at https://gitlab.freedesktop.org/okias/mesa/-/jobs/37478202 at least 10 seconds is needed
17:02 DavidHeidelberg[m]: enunes: I have very elegant fix :P
18:19 DavidHeidelberg[m]: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21711
18:43 alyssa: ERROR - Piglit error: piglit: error: Failed to create waffle_context for OpenGL 3.2 Core Context
18:43 alyssa: ERROR - Piglit error: piglit: error: waffle_context_create failed due to WAFFLE_ERROR_UNKNOWN: eglCreateContext failed with error EGL_BAD_MATCH(0x3009)
18:43 alyssa: but... I don't support GL3.2 ... why are you even trying?!
18:45 HdkR: Easy fix, just support GL 3.2 :P
18:46 alyssa: HdkR: Yeah yeah working on it
19:21 mi6x3m: hey, quick Vulkan question: is it normal to have duplicated indices in the device_compaction_table?
19:22 mi6x3m: like 222,222,223,223,224,224
19:22 mi6x3m: but only for some functions
19:48 mi6x3m: could someone explain what #pragma comment(linker, "/alternatename:_anv_hasvk_CreateCommandPool@104=_vk_entrypoint_stub@0" vk_entrypoint stub means here
19:49 mi6x3m: my entry point anv_hasvk_CreateCommandPool is NULL
19:49 mi6x3m: (I know it's anv_CreateCommandPool in reality), i did some renaming
22:21 jenatali: mi6x3m: Those pragmas are for MSVC
22:21 jenatali: alyssa: Doesn't that error then result in a test skip?
23:04 mi6x3m: Friends please, I'm losing my mind over this literally. I have a segfault with intel-hasvk on https://github.com/krh/vkcube because the method 'vkCreateCommandPool' is NULL
23:05 mi6x3m: I went through the driver's code and the common entrypoints are never initialized so vk_common_CreateCommandPool is NOT set in te dispatch table
23:05 mi6x3m: could anyone confirm this and tell me i'm not going insane?
23:07 mi6x3m: the radv drivers does call add_entrypoints(&b, &vk_common_device_entrypoints, so maybe the intel drivers are missing that??
23:08 kisak: ^dumb question: have you fixed a bug report on the issue tracker?
23:08 kisak: s/fixed/filed/
23:09 mi6x3m: i'm not sure this is a bug, maybe i'm doing something stupid, I've been looking for 3 days now and starting to go insane
23:17 FLHerne: so file a bug and let someone else go insane :p
23:19 FLHerne: hasvk is still pretty new, shallow bugs aren't out of the question
23:19 mi6x3m: as far as im not mistaken the same 'bug' exists in intel as well
23:19 mi6x3m: maybe something went wrong during the split
23:19 mi6x3m: lemme file a bug indeed
23:22 Sachiel: hasvk works fine for me
23:22 Sachiel: so does anv
23:23 mi6x3m: with something that uses vkCreateCommandPool?
23:23 Sachiel: vkcube works fine on hasvk
23:23 Sachiel: and pretty much anything using vulkan uses vkCreateCommandPool
23:24 mohamexiety: hey! sorry if this is the wrong place but I am having a bit of a weird issue. I am trying to clone the nouveau fork of the mesa repo but I can't seem to get over 30 kb/s, and then it just dies after a while. I asked around a few people I know from RADV and neither of them had issues so I wonder if there's something wrong on my end. (I checked my internet connection -- all is fine. also, cloning other non freedesktop repos gives me regular internet
23:24 mohamexiety: speed)
23:25 mi6x3m: Sachiel, well i must be getting mad because i dont see how it can happen without the common_entrypoints
23:25 mi6x3m: which as far as I see are never set in hasvk
23:27 mi6x3m: wait, stop, i think i see it now, it happens through vk_device_init
23:27 Sachiel: vulkan_hasvk/anv_device.c, anv_CreateDevice() calls the vk_device_dispatch_table_from_entrypoints() that should be populating those tables
23:27 mi6x3m: no its slightly different
23:29 Sachiel: what exactly are you doing to run into the problem?
23:34 mi6x3m: Sachiel, I'm working on a frankenstein mod of vulkan :)
23:34 mi6x3m: that... might be released shortly to much outrage
23:34 Sachiel: a what?
23:34 Sachiel: I mean, what steps do you follow to run vkcube and have it crash
23:34 mi6x3m: a very unorthodox custom build
23:35 mi6x3m: well im just building my patched mesa and linking vkcube against it
23:35 Sachiel: so no vulkan loader?
23:35 mi6x3m: with vulkan loader
23:40 Sachiel: without the patches everything works fine?
23:42 mi6x3m: I don't know but I guess so, you say it does and I trust you a lot :)
23:42 mi6x3m: and I just got a suspicion
23:42 mi6x3m: thanks btw, you've been helping me greatly
23:48 mi6x3m: Sachiel, I think what's happening here is that the compiler is silently optimizing vk_common_CreateCommandPool away
23:49 mi6x3m: Sachiel, because looking at the code, no struct members are ever used directly. Instead Mesa converts the table to an array of functions and accesses this way
23:50 Sachiel: lto?
23:51 mi6x3m: yep
23:51 mi6x3m: no struct members ever accessed directly aka all entry points are disregarded
23:51 mi6x3m: how weird
23:52 Sachiel: not sure how well tested lto is, I'd disable it