With the move of the Error Trait into core, the std flag is no longer needed.

[juliankrieger]

I've opened a PR here: #101

With Rust 1.81, the error trait now lives in core. I think coset can live with no_std and alloc only if we reflect that change.

[daviddrysdale]

Thanks!

The code change looks fine, but there would need to be some knock-on effects:

• Presumably this makes 1.81 the MSRV for the crate? (Currently it's 1.58)
• Removal of a feature is going to need a major version bump.

So we'll probably wait for a while before merging this PR.

[juliankrieger]

There are a couple of strategies you could employ to release this earlier.

• You could feature gate the trait impl behind a rust-1.81 feature
• You could feature gate the trait impl behind a v.1.81 specific feature from https://crates.io/crates/rustversion
• Re-export the Error Trait and change the MRSV to 1.81 only for no_std environments
• Introduce a rustc flag similar to anyhow (see core::error, no_std dtolnay/thiserror#304 (comment))

Meanwhile, you could avoid breaking changes by keeping the std feature in the crate, but not apply any cfg attributes that depend on it. I actually use this strategy for all my no_std-by-default crates so that I can avoid a breaking change in the future if I need something from std.

Of course, waiting for Rust v.1.81 to mature is a fine strategy as well. With https://github.com/esp-rs/rust being the only active Rust fork I know also recently hitting 1.81, I would be very interested in hearing your opinion on why keeping the MRSV that low is useful!