On Sat Jan 25 05:43:24 2025 +0000, Paul Gofman wrote:
Why this statistics matters in practice? For non-alertable 0ms waits it won’t yield most of the time and return the corresponding status. For each 0ms alertable wait it *does yield 100% of the time*, do you have any practical reason to lie about that in return status, any app behaves better this way for some explained reason?
I'm following you. Since we have no real control over the scheduler, I guess the second best thing we can do is to just report the fact that any call to `server_select` (from `server_wait`) is an implicit yield (i.e. `SUCCESS`/`USER_APC`). I don't think it's worth arguing over really, maybe your suggestion is safer.
I'll change it in accordance to your suggestion; nevertheless, I still think it should be a `todo_wine` at least, for alertable zero-timeout delays to result in at least some `STATUS_NO_YIELD_PERFORMED`.