Error reporting in scoped tasks

[jplatte]

Hello, and thanks a lot for Rayon!

This issue is more like a question, but GitHub discussions aren't enabled here and I didn't want to use a synchronous communication medium.

I'm currently using scoped tasks (Scope::spawn) for the first time and am wondering if there are any established patterns for reporting errors that happen in those. I think I want to collect all the errors and print them after the scope ends. I guess is easy enough with an MPSC channel? That's what I will try first in any case. I'm just wondering: Are there any established error-handling patterns for scoped tasks? Is this discussed in the documentation somewhere and I just missed it?

[cuviper]

GitHub discussions aren't enabled here

OK, I've flipped that switch so we can try it out!

To the actual question -- I'm not aware of any particular patterns for this, but I think a channel should be fine, depending on how you handle the receiving end. For example, if you block on recv() directly within the scope closure, then that rayon thread will be stuck on this task alone. Maybe you could try_recv() and then yield(), but that's a busy loop chewing time if there's nothing else to be done. Or you could just recv everything after the scope has ended.

[jplatte]

Or you could just recv everything after the scope has ended.

Yeah, that's what I implemented, use a channel with no limit on buffer size and only look at it after all tasks have ended. That means no short circuiting but it's fine for my use case.