Once is not panic-robust about spin-rs HOT 4 CLOSED

Thanks for the bug report 👍
I think this should solve it.

from spin-rs.

There's one design point worth discussing here. Since the Once is not mutated before the builder properly returns, effectively no harm is done to it. So in principle you could reset it back to UNINITIALIZED.

The current behavior is good if most panics inside the builder function are persistent. The behavior I describe is good if most panics inside the builder function are transient. I don't see an obvious benefit to either. What does the standard library do?

from spin-rs.

Ah, the stdlib does what you do:

The closure f will only be executed once if this is called concurrently amongst many threads. If that closure panics, however, then it will poison this Once instance, causing all future invocations of call_once to also panic.

That's fine with me then, thanks!

from spin-rs.

Yes that was what I was imitating.

from spin-rs.