YouTube video by SHADE NET

> Currently you erase all lifetimes (using `'static` or the unstable `'unsafe`) in unsafe code without causing problems, but with EarlyDrop that may result in a use-after-free... Yes, this was something I was a little worried about. Technically we could say that the unsafe code should be guaranteeing that the lifetime hasn't expired (which includes considering `EarlyDrop`). Alternatively, we could make `EarlyDrop` unsafe to make it clear that using it could cause issues for some unsafe code. But that leads to the next issue: > Calling the drop impl before the last value would only be sound if the drop impl doesn't invalidate the safety invariant of the value that gets dropped. I don't understand what this means. What do you mean by "last value"? Isn't this the last borrow's lifetime? Unless you're referring to unsafe code, in which case I argue that your unsafe block isn't guaranteeing the objects lifetime. To be clear, `EarlyDrop` would likely need to occur over an edition boundary since it can cause unsafe code to be unsound.

DaveTheSheep: > `EarlyDrop` > > * You can already drop a variable early (moving it, or `let _ = ...`) > * We can track borrows to a variable and when those lifetimes end (since we know when they overlap). > * Borrows sort of already do this > Another disadvantage of this is that codegen will depend on the exact lifetimes inferred by borrowck. Currently you erase all lifetimes (using `'static` or the unstable `'unsafe`) in unsafe code without causing problems, but with EarlyDrop that may result in a use-after-free if borrowck thought the last use of a variable is earlier than it actually is.

CTP S1E49 SHOW NOTES ( listen (Sat May 25 2024 and thereafter) at: http://buzzsprout.com/2210487 )... ChristiTutionalist Politics (S1E49) "Mother's Day (retrospective)" (Pre-Recorded on Mother's Day...