Enumerations: how are they stored and other questions

[quinedot]

@simonbuchan

Sure, but ! is the canonical uninhabited type, the whole ABI is "there isn't one". How do you have the "same ABI" as that, but it does exist?

Pick an ABI that doesn't rely on inhabitedness. For ! specifically, that would be:

• Any two types with size 0 and alignment 1 are ABI-compatible.

(I'm not aware of any specific guarantees for uninhabited types that don't meet those requirements, but the compiler just needs to be consistent.)

@MichaelV

so I see no reason why any of this should be allowed if T is uninhabited

I think the motivation for MaybeUninit explains this better than I could summarize.

The alternatives tend to not play great with the type system (magic one-off functions, post-mono errors, new auto-traits that potentially also infect and break existing code, ...). There's a lot more conversation about those in the PR thread too.