05:00DavidHeidelberg: zmike: I have feeling like that I have GL-CTS without my patch.. but it's applied a top of same version as mesa uses
11:48Company: MrCooper: GDK_BACKEND=x11 GDK_DEBUG=vulkan GSK_RENDERER=vulkan gtk4-demo (or any other GTK app) on XWayland apparently returns VK_SUBOPTIMAL_KHR - and I wonder if there's something deeper going slightly wrong in either GTK or X or Mesa but I have no good idea how to investigate that
11:49Company: afaiu I get that when the swapchain size doesn't match the window size, but that should never happen because GTK recreates the swapchain on resizes
11:49Company: but it might be something like AcquireNextImage() seeing a resize before GTK does or something?
11:50Company: yet, why would that happen?
11:52zamundaaa[m]: Ugh, stupid matrix bridge, always detecting I was kicked out after posting a comment
11:53zamundaaa[m]: Company: maybe it could be dmabuf feedback, triggering a reallocation of buffers that way?
11:53Company: it happens at the 6th frame after startup, so no idea what's going on
11:55Company: apparently it blocks because all 4 buffers are busy?
11:55Company: huh?
11:56Company: context is from https://gitlab.gnome.org/GNOME/gtk/-/merge_requests/7204#note_2102010 btw
12:04Company:decides to poke MrCooper about it on Monday
12:30zmike: DavidHeidelberg: not sure what you mean
12:31zmike: didn't that cts patch go in last year?
15:36daniels: Company: it happens because you could potentially be scanning out if you reallocated
15:36daniels: whenever you see SUBOPTIMAL, create a new swapchain with pOldSwapchain set accordingly
16:01Company: daniels: it just feels off that it's so easy to reproduce, like if there's something more serious hidden behind it
16:27DavidHeidelberg: zmike: the patch went in like 2 months ago and for version used by MesaCI has to be backported until new release is out
18:46zmike: DavidHeidelberg: you're supposed to tell me these things 🤕
18:46zmike: who do you think it is that causes cts releases to happen
18:49DavidHeidelberg: I backported the patch to the MesaCI so that's fine, but locally I'm having the problems :(
18:49DavidHeidelberg: and Intel CI has the problems
18:50DavidHeidelberg: (but they backported the patch too)
19:39zmike: you're saying you have issues locally using cts from khronos main?
19:40DavidHeidelberg: One time I didn't, now I have..
20:48DavidHeidelberg: karolherbst: I think I seen something like "mold: error: undefined symbol: __asan_stack_malloc_8" with tests + LTO build when testing other MR
21:18karolherbst: mhhh
21:18karolherbst: maybe mold broke it...
21:35DavidHeidelberg: karolherbst: before the mold update thou 😉
21:35karolherbst: oh no
22:33Lynne: amd again at it with gfxXX codenames
22:33Lynne: when they themselves get confused months later and want to change it to naviXX