Taiki Endo

@pkubaj Hmm... that's odd. https://github.com/crossbeam-rs/crossbeam/pull/698 was supposed to have taken care of those problems. Which build system did you use? And did you receive any warnings during the build?

If we handle this on our end, it seems increasing PINNINGS_BETWEEN_COLLECT or to make it configurable is okay. Which should be chosen depends on what the adverse effects of increasing...

> * Adjust rayon to use separate `Global`s for each thread pools as a kind of scope? cc @cuviper: Any thoughts on this idea (or this issue)?

I would accept a PR that does this in a similar way as https://github.com/crossbeam-rs/crossbeam/pull/487.

FYI: I tried an implementation based on the single-element queue used in concurrent-queue (async-channel): https://github.com/crossbeam-rs/crossbeam/commit/77e70aed60c4b98c94f6e6f227497a3f856573b1 The performance of oneshot and SPSC has been improved, but there were regressions in the...

BTW, in addition to this, I'm considering porting the 's changes to crossbeam.

@m-ou-se Thanks! I pushed [a fix](https://github.com/crossbeam-rs/crossbeam/compare/1aca853b7a81881410ff8196d01fa459fd37d330..e637a68d020df7589a781041608ba0d1736e4716) for that bug EDIT: The fix is: - If the closure is already finished when the handle is dropped, drop the result. - If...

bors r+

bors r- miri failed due to thread leaks: https://github.com/crossbeam-rs/crossbeam/runs/7479452681?check_suite_focus=true

This is because `crossbeam::scope` is defined in `crossbeam_utils::thread` and re-exported as `crossbeam::scope`. https://github.com/crossbeam-rs/crossbeam/blob/2f1ed42da69f93d54d4a648c1fee6c54fb739aec/src/lib.rs#L101 Ideally, I would like to adjust the document according to the current crate, but this is currently...