Add Channel#closeWithElement

[tothpeti]

This PR adds the requested functionality: #3279

• Add closeWithElement in Channel
• Add test case

[armanbilge]

More name ideas: sendThenClose or sendLast.

[tothpeti]

More name ideas: sendThenClose or sendLast.

I think sendAndClose would be pretty clear for the user.

[SystemFw]

I've been trying to figure out the correct semantics for this operation, which I think depends on expected use case, because as it stands, sending is not atomic with closing with this implementation.
In particular cancelling a sendAndClose blocked on a bound means the channel is closed without having the element sent included in the channel's graceful closure.

[SystemFw]

If I understand the use case correctly, I think the correct behaviour for this operation is to be called closeWithElement, and bypass the bound on a bounded channel without ever blocking, but enqueueing the final element alongside setting closed =true.

[armanbilge]

Thanks. That sounds right to me, @tothpeti wdyt?

[tothpeti]

Thanks. That sounds right to me, @tothpeti wdyt?

Sounds good to me, as well.

[armanbilge]

Hmm, I don't think this case should ever happen? Is it happening?

[tothpeti]

It is happening once or twice during 10000 runs, but it is pretty rare.

[armanbilge]

I don't see how that can happen if everything is working correctly. If the closeWithElement returns that the Channel is already closed, then who is closing it? Seems like a bug 😕

[tothpeti]

To be honest, I left this match case there, when we iterated over multiple possible solutions. So, probably this will not happen. I will try it without this match case soon.

[tothpeti]

Yep, I tried with 100000 runs, and that match case never occurred. So, it can be safely removed. 👍

[armanbilge]

Looks good. Thanks for all your work on this!