scottmcm

Spitballing: what if `Foo { .. }` default-initialized even the things that *don't* have defaults? (Basically, assuming we have a `const Default` of some sort, it puts in `Foo {...

The decision to require the `else` to be diverging was made very intentionally. It'll take much more than "but it should" to change that. I'm going to close this and...

An idea for an alternative way to approach this: https://internals.rust-lang.org/t/idea-diagnostic-aliases-attribute/13366?u=scottmcm

@rfcbot concern ambiguity This makes things like ```rust type transparent = u64; #[repr(transparent)] enum Foo { … } ``` legal. That seems particularly bad for any other proc macros that...

@rfcbot resolve ambiguity I'm an idiot and somehow missed the "you need `self::` part" 🤦 That does solve the hard blocker, but it still makes me wish for a better...

I went to mark this waiting-on-team, but it looks like that label doesn't exist here. Please treat this RFC as waiting on us to give better feedback.

I think `r=me` for the code change here, since it looks like tests pass. That said, I could have sworn there was more that would have been needed here. If...

> Should it not instead shrink to the smallest possible power of two capacity? I don't think so. If I happen to have a capacity-257 vector, I don't want it...

> I think users interested in using this function would want to control the strategy or load factor rather than trust the implementation choice. I don't think so, because if...

By "quadratic" in the description I meant that the overall algorithm would be quadratic, not `shrink` itself, like how using `reserve_exact(1)` makes a push loop become quadratic even though each...