13:11 <MoeIcenowy> https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/gallium/drivers/panfrost/pan_resource.c#L967 is this something that never happens or is it a TODO?

13:13 <alyssa> MoeIcenowy: I have no idea if it ever happens, but it doesn't matter

13:13 <MoeIcenowy> alyssa: I met glMapBufferRange() getting NULL in KWin

13:13 <MoeIcenowy> (although I'm not sure whether is it related to this

13:14 <alyssa> We only bail with PIPE_MAP_DIRECTLY (in conjunction with other options), and Gallium's contract says if transfer_map() returns NULL with PIPE_MAP_DIRECTLY, the frontend must retry without PIPE_MAP_DIRECTLY

13:14 <alyssa> (...I think)

13:19 <MoeIcenowy> well I see no retry in https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/mesa/state_tracker/st_cb_bufferobjects.c#L556

13:21 <alyssa> Hm.

13:21 <MoeIcenowy> well it looks that no PIPE_MAP_DIRECTLY will appear here?

13:27 <MoeIcenowy> maybe I should just debug from kwin side... how can I debug why a NULL return value is get in application?

13:35 <MoeIcenowy> BTW src/panfrost/vulkan/panvk_device.c:87:13: warning: ‘mesa_timestamp’ is used uninitialized in this function [-Wuninitialized]