Nemo157

@rushmorem I'm a bit confused about your example ```rust #[async] fn fetch_rust_lang(client: hyper::Client) -> impl Future { // We are forced to await here even though we don't have to....

Another thought I just had, this should also allow easy integration of nominal types once the new `impl Trait` type aliases are available. Something like ```rust type RustLang = impl...

I finally got round to pushing an updated branch implementing this: [Nemo157#alternate-signatures-2](https://github.com/Nemo157/futures-await/tree/alternate-signatures-2). Including tests showing that it fixes #11/#53 via the normal lifetime syntax. It currently has some pretty terrible...

@withoutboats I actually like that part of this proposal because it makes `#[async]` behave more like other Rust code, if I were to see ```rust #[async] fn foo(&self, arg: &str)...

Yep, I was misrembering how lifetime elision works. Still, I think that improving how `impl Trait` handles lifetimes would be a better path than hiding the lifetimes away in the...

Referring back to my [old comment about nominal `impl Future` types](https://github.com/alexcrichton/futures-await/issues/15#issuecomment-337197156), I just accidently attempted to use this today, so it's definitely something that would be useful (once abstract/existential types...

Only used at build-time through syntect ``` yaml-rust v0.4.5 └── syntect v5.2.0 [build-dependencies] └── docs-rs v0.6.0 ```

It's not just shrinking, allocators that overallocate (as allowed by the `Allocator::allocate` docs) also fail under miri: ```rust #![feature(allocator_api)] use core::ptr::NonNull; use std::alloc::{Allocator, Layout, AllocError, System}; struct OverAllocate; unsafe impl...

By "overallocating" I mean allocating something larger than the requested layout _and returning the entire larger allocation_, not returning a subset.

Does whitespace impact the parser speed much? I assume this will make very little transfer size difference post-compression.