On Sat Jan 25 05:37:01 2025 +0000, Paul Gofman wrote:
I think matching return status statistics by faking status is, most of the time, is the wrong metric. The status should reflect the real state of things, whether yield performed or not. Then, if anything cares about it at all, we shouldn’t probably err out on the side saying that yield wasn’t performed, it msy trigger unneeded waits.
I mean, before this, we were returning a completely unexpected result for `NtDelayExecution` (`STATUS_TIMEOUT`). If, by your suggestion, we instead always return `STATUS_SUCCESS` for alertable waits, that'd be almost the complete inverse of what Windows returns (for 1ms resolution). If an application happened to care about the return value of the delay, I feel like the current version is more likely to be correct than if we basically flipped the NO_YIELD/SUCCESS ratio from ~100%/~0% to 0%/100%.