Is there a meeting today?
My calendar has an event for just now.
I suppose not. For anyone who shows up later, here's my update for this past week:
@nikomatsakis did you have a chance to look into the lifetime issues on friday?
I plan to open a PR with only the MIR/analysis changes so we can discuss
Sorry I've been ill, but better now. Shrank the problem to a smaller repro - I think the question is firstly is the span_bug! definitely a span_bug in this case as it's an existential type so maybe you wouldn't expect the type to be defined at that point?
One thing I am very conscious of is that once we ship async await the very next thing people will be asking for / expecting is streams created by some form of async/generator/yield construct. At the moment what we're shipping supports Future but as far as I can see doesn't really support Stream.
@Giles Cope It's fairly easy to add support for Stream. the issue is that a proper definition of the
stream trait is one which requires GATs
It's actually just a small handful of lines to make the existing system create asynchronous generators that implement the current Stream trait
it's a lot more work to make it support a hypothetical Stream + borrowing
Yep GAT looks non-trivial. I guess releasing non-borrowable async streams probably isn't a good idea incase the implications of the borrowed async streams might take us in a different direction and we ended up with two ways of doing things...