Replying to @chandlerc1024
I like "unformed" here -- seems close to minimum requirements of "partially formed". I don't see rationale for requiring support for multiple destructions, though. Won't that often require users branching in destructors when not vacuous? Is rationale for that aspect up anywhere?
2
1
The specific goal is to represent the large body of types which have a viable idempotent state for destructors. As for how multiple destruction comes up, some of it may be in the proposal. My memory is that it is only about reducing cost -- specifically dynamic tracking.
Load more