Lewis Baker

My main concern is that what is currently proposed in P2300R0 does not allow customisation by defining member functions and yet we are doing that in libunifex which makes it...

Thanks for the suggestion. The intrusive heap structure is indeed rather naive and placeholder at the moment and won't scale well to large numbers of timers, but was simple to...

The `task` coroutine type does not yet support transparent propagation of execution context/scheduler to its child operations so your example's usage of `current_scheduler` inside the coroutine won't work. This is...

libunifex is only really using the core syscall parts of io_uring and isn't making use of any of the liburing helpers. However, I have had difficulty trying to just include...

Is the idea here to allow wrapping a sender with something that uses an async-scope and then having it detach-on-cancellation and immediately complete with set_done but still have the async-scope...

> is this same as https://github.com/facebookexperimental/libunifex/blob/master/source/win32/windows_thread_pool.cpp The windows_thread_pool implementation currently only supports the scheduler and time_scheduler interfaces (`schedule()`, `schedule_at()`, `schedule_after()`, `now()`). This issue is about extending that implementation with support...

Yes, I can see how this would be problematic for some use-cases. We can probably be somewhat less aggressive at decaying lvalues in the `when_all` algorithm. There can be difficulties...

PR #44 changes on() to take a scheduler. Still need to tackle via().

PR #47 changes `typed_via()` to take a scheduler.

There are some tradeoffs with its behaviour here with launching the source first or the trigger first. One thing I've been thinking about is actually having it launch the trigger...