NIOAsyncWriter: Fix suspending yield when writer is finished

[orobio]

This fixes hitting the following preconditionFailure in NIOAsyncWriter: "This should have already been handled by yield()".

It doesn't expect a yield to be suspended when the state is .writerFinished, but this can definitely happen. This seemed to be the correct solution to me, but please check it carefully, because I'm unfamiliar with this code.

I'm not happy about the test for this. It's blocking the didYield call for a while, to make sure the correct state is reached. See also the 'FIXME' line. What would be a better way to do this?

[Lukasa]

This is a great catch, thanks so much!

Regarding your test, it might be useful for the test to use a new delegate that makes it easier to arrange the ordering to work out. NIO's ConditionLock can be used to set up a nice ordering chain, so that instead of using the usleeps we can just block on that lock being unlocked with the right value.

[orobio]

@Lukasa : I updated it with ConditionLock. Is this what you had in mind?

This is still blocking a thread from the concurrency pool and then depending on another thread to unblock it. Will that not be a problem? The only way I see around that is to add an async hook to NIOAsyncWriter.

[Lukasa]

For testing purposes, that blocking will be safe. We will eventually resolve it, and we don't rely on Dispatch to make that happen (it requires the NIO thread to do it), so it'll be ok.

[Lukasa]

This is a really lovely patch, thanks so much @orobio!

[orobio]

For testing purposes, that blocking will be safe. We will eventually resolve it, and we don't rely on Dispatch to make that happen (it requires the NIO thread to do it), so it'll be ok.

I'm not sure I fully understand what you're saying here, but I assume it's fine then.

I do still have that 'FIXME' line in there, which would probably be good to remove. Well..., this got merged while I was typing, so I'll open a new pull request for that.

Edit: Opened #3059