feat: add support for compio runtime

[George-Miao]

• feat: add compio support
• feat: add unit test

Checklist

• Updated guide with pertinent info (may not always apply).
• Squash down commits to one or two logical commits which clearly describe the work you've done.
• Unittest is a friend:)

---

This change is 

[drmingdrmer]

Thank you for this supplementary runtime support!

There should be a CI to ensure it pass the test, similar to:
https://github.com/datafuselabs/openraft/blob/00771c8db3dd274fbc2c7461e56960602b94931a/.github/workflows/ci.yaml#L270

Reviewed 10 of 10 files at r1, all commit messages.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @George-Miao)

---

rt-compio/src/lib.rs line 44 at r1 (raw file):

    pub fn cancel(mut self) {
        // SAFETY: We are not using the JoinHandle anymore, so we can safely drop it.
        unsafe { ManuallyDrop::drop(&mut self.0) }

Since tasks are cancelled when their JoinHandle is dropped, spawned compio tasks should use detach() to keep them running in the background:

https://docs.rs/compio-runtime/0.7.0/compio_runtime/struct.Task.html#method.detach

OpenRaft only needs background tasks that run to completion without cancellation, so all spawned tasks can be detached. This eliminates the need for ManuallyDrop wrappers around task handles.

Code quote:

    pub fn cancel(mut self) {
        // SAFETY: We are not using the JoinHandle anymore, so we can safely drop it.
        unsafe { ManuallyDrop::drop(&mut self.0) }

---

Cargo.toml line 63 at r1 (raw file):

    "tests",
    "stores/memstore",
    "rt-compio"

A runtime does not have to be a member. Just put it in the following exclude section.

[George-Miao]

Since tasks are cancelled when their JoinHandle is dropped, spawned compio tasks should use detach() to keep them running in the background:

The thing is, detach is a consuming function, meaning the handle will not be usable after detached, and the trait requires the handle to implement Future so the result can be retrieved after completion.

[drmingdrmer]

Since tasks are cancelled when their JoinHandle is dropped, spawned compio tasks should use detach() to keep them running in the background:

The thing is, detach is a consuming function, meaning the handle will not be usable after detached, and the trait requires the handle to implement Future so the result can be retrieved after completion.

Hmm... I missed this point.

Then it seems when CompioJoinHandle is dropped, it should detach the inner task. Using ManuallyDrop just leaks the inner Task if the user does not call cancel

[George-Miao]

Then it seems when CompioJoinHandle is dropped, it should detach the inner task

That works too. I just need to wrap it with an Option.

[drmingdrmer]

Reviewed 1 of 1 files at r2, 1 of 1 files at r3, 1 of 1 files at r4, all commit messages.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @George-Miao)

---

rt-compio/src/lib.rs line 46 at r4 (raw file):

            task.cancel();
        }
    }

Why does it need a cancel API? OpenRaft does not need to cancel a task.

If it allows user to cancel it, the following poll() will just panic. Without it, everything looks good, right?

Code quote:

    pub fn cancel(mut self) {
        if let Some(task) = self.0.take() {
            task.cancel();
        }
    }

---

rt-compio/src/lib.rs line 73 at r4 (raw file):

impl<T> FusedFuture for CompioJoinHandle<T> {
    fn is_terminated(&self) -> bool {
        self.0.is_none()

Without cancel(), it does not need to be fused any more.

Did I miss anything?

Code quote:

impl<T> FusedFuture for CompioJoinHandle<T> {
    fn is_terminated(&self) -> bool {
        self.0.is_none()

[George-Miao]

The internal handle is pub, so everyone can access and alter the state. I can remove all these and hide inner to users if you feel ok.

[drmingdrmer]

The internal handle is pub, so everyone can access and alter the state. I can remove all these and hide inner to users if you feel ok.

Yes, it would be better. There is no need to leak internal non-required features to users.

One more public features just introduces additional maintenance burden.

[drmingdrmer]

LGTM!

Reviewed 1 of 1 files at r6, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @George-Miao)

[George-Miao]

There was a problem in the code, fixed.

[drmingdrmer]

Reviewed 1 of 1 files at r8, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @George-Miao)

[drmingdrmer]

Reviewed 1 of 1 files at r10, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @George-Miao)