Taiki Endo

I would like to switch to `Once` or blocking method of async `OnceCell` recently added to async-lock, etc. if possible. However, if it is difficult, let's bump MSRV. The approach...

Sounds good to me.

At this time, I tend to prefer the third (unix only). (Eventually, we probably need to implement them on windows, but there is no need to rush...)

As for the second issue, it is the same issue as https://github.com/rust-lang/rust/issues/81684. One or both of the following may work: - support [minicov](https://github.com/Amanieu/minicov) (but, I guess we can already handle...

You need to use --ignore-unknown-features flag https://github.com/taiki-e/cargo-hack#--ignore-unknown-features > Skip passing `--features` to `cargo` if that feature does not exist.

It makes sense to mention that flush is needed. [BufWriter](https://docs.rs/futures-lite/latest/futures_lite/io/struct.BufWriter.html) also has similar docs.

I have not investigated the details regarding alternatives, but at this point I think it is reasonable to improve the docs. As for std scope threads, it is usually incompatible...

According to https://github.com/rust-lang/rust/pull/93563 by @ibraheemdev, is the real problem here not the unbounded spinning in unbounded queue (i.e., wait_write pointed by #366), but the amount of spin_loop_hint and yield_now and...

yeah, this is ready to review, as PR is not marked as a draft.

Argument support is a feature that was added intentionally. (see https://github.com/tokio-rs/tokio/issues/1564 and https://github.com/tokio-rs/tokio/pull/1594) My comment in https://github.com/tokio-rs/tokio/pull/3766 is about the `self` argument. (And since #3766 I've actually [started using `self`...