Perform spurious wakeups from `park`?

[RalfJung]

The specification of park allows spurious wakeups. I am not entirely sure what our implementation of the underlying OS primitive currently entails, but it might be that it will never spuriously wake up. In which case it'd be nice if we could somehow artificially make it wake up -- like with compare_exchange_weak, that is probably a not-well-tested code path in programs using park.

[bjorn3]

At least park_timeout can have spurious wakeups in practice on linux as there is no way to distinguish a spurious wakeup and a timeout: https://github.com/rust-lang/rust/blob/baaa3b682986879c7784b5733ecea942e9ae7de3/library/std/src/sys_common/thread_parker/futex.rs#L78-L80 On macOS park_timeout will spuriously wakeup after 1000 years to work around an OS bug: https://github.com/rust-lang/rust/blob/baaa3b682986879c7784b5733ecea942e9ae7de3/library/std/src/sys/unix/thread_parker.rs#L64-L66 On windows park_timeout can also spuriously wakeup for the same reason as linux: https://github.com/rust-lang/rust/blob/baaa3b682986879c7784b5733ecea942e9ae7de3/library/std/src/sys/windows/thread_parker.rs#L158-L160

[RalfJung]

Right, but can that actually happen with the Miri implementation of futex_wait?