No overflow panic on very long timeouts #111
Conversation
src/condvar.rs
Outdated
| #[cfg(feature = "nightly")] | ||
| let very_long_timeout = Duration::from_secs(u64::max_value()); | ||
| #[cfg(not(feature = "nightly"))] | ||
| let very_long_timeout = Duration::from_millis(u32::max_value() as u64); |
There was a problem hiding this comment.
This would look better with if cfg!(feature = "nightly") {} else {}.
There was a problem hiding this comment.
I instinctively always try to use #[cfg] to make the unwanted code go away as early as possible in the compile :)
I also think that won't work. Since checked_add does not exist on non-nightly. So it does not compile. if false { checked_add() } still requires checked_add even if it's never going to be executed.
There was a problem hiding this comment.
Sorry.. I misread which file the comment was on. Yes, here we can use cfg!. Fixed it.
|
The Linux CI error seems to be because Some of the windows errors seems to be because So maybe we need to bump the minimum version here anyway? Or lock these dependencies to very specific older patch releases. Requiring a newer compiler sounds very much like a breaking change to me, so I don't understand why these crates didn't bump the major version, but anyway. |
faern
force-pushed
the
no-instant-overflow-panic
branch
from
fd53e62
to
ad5dc16
Compare
faern
force-pushed
the
no-instant-overflow-panic
branch
from
ad5dc16
to
e593f10
Compare
Unstable support for checked_add on time types was finally merged (rust-lang/rust#56490) so now we can avoid the possible panic that is
Instant::now() + timeout.First commit is unrelated. It's just some other simplification I thought it was nice to make.