20:57 <eric_engestrom> kusma: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/panfrost/lib/kmod/panthor_kmod.c#L1141 looks like a bug to me, having drmSyncobjTimelineWait() inside the assert() means the behaviour of the function depends on the buildtype

20:58 <eric_engestrom> (I would tag boris as well but I don't think he's online)

21:01 <eric_engestrom> (since you two are the author and reviewer on the commit that introduced that code)

21:46 <daniels> eric_engestrom: it’s not a load-bearing wait; it really is just a defensive check and works fine without it

21:46 <eric_engestrom> ack, glad to hear :)

21:47 <eric_engestrom> it does mean the value of `new_sync_point` set varies though

21:48 <eric_engestrom> but I guess that's not a problem?

21:51 <daniels> timeline syncobjs can contain a bunch of different fences, and this is basically validating that the timeline value we’ve just claimed to commit, really has been sent to hw

21:56 <eric_engestrom> ack

21:56 <eric_engestrom> so no bug, just code that's not meant to be read by someone who doesn't know what they're reading :P

22:12 <kusma> eric_engestrom: if it helps, I've complained about the same in the past :P

22:13 <eric_engestrom> ^^