06:49tzimmermann: jfalempe, thanks for reviewing the panic-helper series. i currently sort out the comments that I got. do you know if some driver provides a panic display outside of primary planes?
07:00MrCooper: zamundaaa[m]: seems more like system mis-configuration to me
07:01MrCooper: zamundaaa[m]: BTW, since the client doesn't know the duration of vblank, Mesa can't set commit-timing 500 µs before start of vblank, maybe rather before end of vblank?
07:03MrCooper: also, without specific support in the KMS UAPI, handling it in the compositor still has a similar issue as handling it in the client, just less likely to do the wrong thing
07:04MrCooper: also, there are relevant HW limitations, e.g. (some?) Intel GPUs can't switch between sync & async flips arbitrarily
07:05MrCooper: all in all it's a surprisingly complex topic, and since VRR is the better solution, I'm finding it hard to care too much
09:43zamundaaa[m]: MrCooper: yeah, it's 500us before the predicted presentation timestamp
09:45zamundaaa[m]: And true, perfect support would require hardware to handle it