Re: [PATCH v9 7/8] rust: Add read_poll_timeout functions

[Gary Guo <gary@garyguo.net>]

On Mon, 27 Jan 2025 15:31:47 +0900 (JST)
FUJITA Tomonori <fujita.tomonori@gmail.com> wrote:

> On Mon, 27 Jan 2025 11:46:46 +0800
> Gary Guo <gary@garyguo.net> wrote:
> 
> >> +#[track_caller]
> >> +pub fn read_poll_timeout<Op, Cond, T: Copy>(
> >> +    mut op: Op,
> >> +    mut cond: Cond,
> >> +    sleep_delta: Delta,
> >> +    timeout_delta: Delta,
> >> +) -> Result<T>
> >> +where
> >> +    Op: FnMut() -> Result<T>,
> >> +    Cond: FnMut(&T) -> bool,
> >> +{
> >> +    let start = Instant::now();
> >> +    let sleep = !sleep_delta.is_zero();
> >> +    let timeout = !timeout_delta.is_zero();
> >> +
> >> +    if sleep {
> >> +        might_sleep(Location::caller());
> >> +    }
> >> +
> >> +    loop {
> >> +        let val = op()?;
> >> +        if cond(&val) {
> >> +            // Unlike the C version, we immediately return.
> >> +            // We know the condition is met so we don't need to check again.
> >> +            return Ok(val);
> >> +        }
> >> +        if timeout && start.elapsed() > timeout_delta {  
> > 
> > Re-reading this again I wonder if this is the desired behaviour? Maybe
> > a timeout of 0 should mean check-once instead of no timeout. The
> > special-casing of 0 makes sense in C but in Rust we should use `None`
> > to mean it instead?  
> 
> It's the behavior of the C version; the comment of this function says:
> 
> * @timeout_us: Timeout in us, 0 means never timeout
> 
> You meant that waiting for a condition without a timeout is generally
> a bad idea? If so, can we simply return EINVAL for zero Delta?
> 

No, I think we should still keep the ability to represent indefinite
wait (no timeout) but we should use `None` to represent this rather
than `Delta::ZERO`.

I know that we use 0 to mean indefinite wait in C, I am saying that
it's not the most intuitive way to represent in Rust.

Intuitively, a timeout of 0 should be closer to a timeout of 1 and thus
should mean "return with ETIMEDOUT immedidately" rather than "wait
forever".

In C since we don't have a very good sum type support, so we
special case 0 to be the special value to represent indefinite wait,
but I don't think we need to repeat this in Rust.