libgreen sched::test_spawn_sched_blocking is unstable #11200
Comments
|
This also appears to have failed on the linux bot recently: http://buildbot.rust-lang.org/builders/auto-linux-32-nopt-t/builds/2350/steps/test/logs/stdio |
flip1995
pushed a commit
to flip1995/rust
that referenced
this issue
Jul 31, 2023
[`unused_async`]: don't lint if paths reference async fn without immediate call Fixes rust-lang#9695 Fixes rust-lang#9359 Clippy shouldn't lint unused `async` if there are paths referencing them if that path isn't the receiver of a function call, because that means that the function might be passed to some other function: ```rs async fn f() {} // No await statements, so unused at this point fn requires_fn_future<F: Future<Output = ()>>(_: fn() -> F) {} requires_fn_future(f); // `f`'s asyncness is actually not unused. ``` (This isn't limited to just passing the function as a parameter to another function, it could also first be stored in a variable and later passed to another function as an argument) This requires delaying the linting until post-crate and collecting path references to local async functions along the way. changelog: [`unused_async`]: don't lint if paths reference async fn that require asyncness
flip1995
pushed a commit
to flip1995/rust
that referenced
this issue
Oct 3, 2024
Remove method call receiver special casing in `unused_async` lint Fixes the false positive mentioned in rust-lang/rust-clippy#13466 (comment). The false negative in the OP would be nice to fix too, but I'd rather do that in a separate PR because it's much more involved Before this change, the `unused_async` lint would check if the async fn is also used anywhere and avoid linting if so. The exception is if the async function is immediately called, because the returned future handling can be easily removed (and also if we don't have some exceptions then the lint wouldn't trigger anywhere) *or* if it's a method call receiver. I'm not exactly sure why I implemented that special casing for method call receivers in rust-lang#11200, but it doesn't make much sense in hindsight imo. Especially given that method calls are essentially equivalent to function calls with the receiver as the first argument, which was the primary motivation for not linting in the first place (async fn passed to another function, like `axum::get(handler)` where handler has to be an async fn). changelog: none
# for free
to join this conversation on GitHub.
Already have an account?
# to comment
Operating with a completely fresh build (
git clean -fdx), the libgreensched::test_spawn_sched_blockingtest is sometimes hanging for me. (The other 24 tests in libgreen are successfully passing or being ignored, as appropriate.)First of all I run
make checkand it hangs at that point. I kill it (^C) and try again and it passes.And it hangs. Kill and retry the
make check-stage2-green, and it passes.As long as I
rm -fthe .ok file before runningmake check-stage2-green—whether or not that file exists—it fails, but if I then runmake check-stage2-greenagain, it works again. The scary part is that this behaviour seems to be consistent, though that's ridiculous.The text was updated successfully, but these errors were encountered: