Release 3.8.0

[taiki-e]

Changes:

• Implement AsRawFd and AsFd for Poller on Redox OS. (Implement AsRawFd for Poller on Redox OS #235)
• Update hermit-abi to v0.5.0. (Update hermit-abi requirement from 0.4.0 to 0.5.0 #229)
• Update rustix to 1.0 (Port to rustix 1.0. #230). This also fixed a bug in wait which (contradicting docs) cleared the events vector instead of appending to it.
• Add support for QNX. (port to qnx #201)

[fogti]

• Ported to rustix 1.0 (Port to rustix 1.0. #230), this also fixes a bug in wait which (contradicting docs) cleared the events vector instead of appending to it.
• Fixed compilation on QNX (port to qnx #201)

[fogti]

Current state of this repo doesn't seem to induce regressions with some simple testing by replacing the polling used in topola-egui on x86_64-gnu-linux at least.

[zeenix]

looks like the CI is failing here due to:

Is it a call to move away from circle-ci?

[taiki-e]

@zeenix It has nothing to do with failure at all, and is something that also shows up in jobs that pass (e.g., see freebsd job).
What is the cause of the failure and how it can be fixed is already explained in my comment linked in #237.

[zeenix]

@zeenix It has nothing to do with failure at all, and is something that also shows up in jobs that pass (e.g., see freebsd job). What is the cause of the failure and how it can be fixed is already explained in my comment linked in #237.

Thanks. Should we remove the job from this project (and others) too? 🤔

[taiki-e]

As I said in async-io PR (smol-rs/async-io#222 (comment)), I'm fine with temporarily disabling OpenBSD test, but I don't want to remove it forever.

We could do the same thing we did in async-io, or we could simply wait for an upstream fix PR, since it now exists (anarazel/pg-vm-images#113).