We have decided to merge this ASAP, so I can start opening the rest of the PRs. This should be fine since the API is not used anywhere yet, but we will do more testing before we merge the other PRs.
sem_timedwait uses an absolute time value, that's why we need the clock_gettime call.
I've changed those redundant CHECKs for ERRORs.
EAGAIN is for sem_trywait.
The equivalent for sem_timedwait is ETIMEDOUT, which is what we check here.
No, because we pass 0 as parameter, which means no wait.
See the beginning of the function: if (wait_ms == 0) {.
Ah, I somehow thought FILE_MAX was the same value as PATH_MAX.
When assert triggers it means something unexpected has happened.
But that's true here too. Those asserts should never trigger. It's only that, since errors here could lock Blender, we try to…
Yes, this was wrong and was breaking the lost process detection. It should be solved now.
Ooops! That's exactly the case, I've used seconds initially. Thanks!
I've removed the CHECKs from conditionals, but I can't say I get your point. Other than asserts triggering or not (which is always the case), I don't see how the behavior between release and debug…