Stream: t-libs/wg-allocators

Topic: Concerns on nightly in alloc-wg?


Tim Diekmann (Nov 21 2019 at 16:38, on Zulip):

As the alloc-wg crates grows, I think it makes sense to require a nightly compiler as the feature list grows and many things cannot be used on stable anyway. The biggest downsides so far are dispatch_from_dyn, coerce_unsized and specialization for Vec. Additionally, it will introduce dancing around core::ptr::Unique, as NonNull does not implement Send or Sync.

Are there any concerns not to require the nightly compiler?

Gurwinder Singh (Nov 21 2019 at 18:07, on Zulip):

I agree with using nightly only. The split in code is starting to get bad

Simon Sapin (Nov 28 2019 at 17:19, on Zulip):

Nightly or not, I would recommend against using Unique outside of the standard library, it’s unclear whether it’ll always exist. Explicit impls of Send and Sync on types that contains NonNull should work, would that be satisfactory? Or make your own private newtype of NonNull that implements Send and Sync as needed.

Tim Diekmann (Nov 28 2019 at 19:05, on Zulip):

Good advice, I'll change that.

Last update: Dec 12 2019 at 00:45UTC