Re: [PATCH weston 1/1] compositor: Abort on bad page flip timestamps

On 07.11.2014 18:36, Pekka Paalanen wrote:

If we still feel that exiting Weston on the first timestamp failure (zero or not increasing) is too harsh, we could simply log a warning and add again some heuristics on when things are too far wrong to continue.

E.g. getting a single timestamp failure during the first 5 seconds of compositor uptime warrants an exit, but later probably not. We want to catch cases where the compositor cannot reliably work to begin with, but killing the user's session much later is pretty blunt if it already worked somewhat.

Would that be good?

Yeah, that would address my main concern.