use par_for_each_in in par_body_owners and collect_crate_mono_items
#99457
Conversation
rustbot
added
the
T-compiler
|
r? @fee1-dead (rust-highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
This comment has been minimized.
This comment has been minimized.
| @@ -146,7 +146,7 @@ cfg_if! { | |||
| t.into_iter() | |||
| } | |||
|
|
|||
| pub fn par_for_each_in<T: IntoIterator>(t: T, for_each: impl Fn(T::Item) + Sync + Send) { | |||
| pub fn par_for_each_in<T: IntoIterator>(t: T, mut for_each: impl FnMut(T::Item) + Sync + Send) { | |||
There was a problem hiding this comment.
Wouldn't this break the parallel compiler? The requirement is relaxed and then if we switch to parallel rustc it would fail to compile.
There was a problem hiding this comment.
It doesn't break the compiler from actual testing.
This function signature change is only required in non-parallel mode, and the one in parallel mode does not need to be changed. I think this is because the internal implementation of IntoParallelIterator does some kind of encapsulation of this(not sure).
#[Edit] See the new explanation below
There was a problem hiding this comment.
If we want to maintain consistency in both modes, we should change Fn in the signature in parallel mode to FnMut. I'm not sure if this is necessary.
There was a problem hiding this comment.
I don't think we can do that, if we run it in parallel then the function would be called from multiple threads thus requiring it to be called without having exclusive access.
There was a problem hiding this comment.
That makes sense. So I think we can keep the Fn in the signature in parallel mode, as it doesn't cause compilation errors
There was a problem hiding this comment.
I revisited the implementation of both patterns. The place to call this interface is in collect.rs. Because the MTRef type is not a &mut reference in parallel mode, the Fn type can be used without causing a compilation error.
|
@bors r+ |
bors
added
S-waiting-on-bors
…askrgr Rollup of 9 pull requests Successful merges: - rust-lang#98028 (Add E0790 as more specific variant of E0283) - rust-lang#99384 (use body's param-env when checking if type needs drop) - rust-lang#99401 (Avoid `Symbol` to `&str` conversions) - rust-lang#99419 (Stabilize `core::task::ready!`) - rust-lang#99435 (Revert "Stabilize $$ in Rust 1.63.0") - rust-lang#99438 (Improve suggestions for `NonZeroT` <- `T` coercion error) - rust-lang#99441 (Update mdbook) - rust-lang#99453 (:arrow_up: rust-analyzer) - rust-lang#99457 (use `par_for_each_in` in `par_body_owners` and `collect_crate_mono_items`) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup
Using
par_iterin non-parallel mode will cause the entire process to abort when any iteration panics. So we can usepar_for_each_ininstead to make the error message consistent with parallel mode. This means that the compiler will output more error messages in some cases. This fixes the following ui tests when setparallel-compiler = true:This refers to #68171
Updates #75760