Breaking Change - encode_utf8 destabilised

[thepowersgang]

The change 2b3c5ac converted char::encode_utf8 from a stable method of an unstable trait to an unstable method.

The trait CharExt is imported as part of libcore's prelude, which can be done on the stable build (see https://play.rust-lang.org/?gist=7320c1273bdc3dde5450&version=stable).
Under current nightly, the same code fails to compile due to this same API.

This is technically a regression, and may need a crater run to check if it breaks anything (unlikely), and may need closer scrutiny.

[thepowersgang]

cc #27784 @alexcrichton

[thepowersgang]

Note, discussion in the PR #32204 on the relevant lines. Appears that breakage was acknowledged and assumed to not have any impact.

[bluss]

It's a regression in the sense that it breaks / changes behavior of stable code (but this doesn't alone imply that, or imply that it not, needs to be fixed.)

[alexcrichton]

triage: I-nominated

nominating for discussion in a libs triage meeting. @thepowersgang is this enough enough to work around for you? For example, do you have crates that build on stable which require this method?

Ideally we could let this slide, but if the breakage is too much we'll just have to bite the bullet and revert.

[thepowersgang]

The only breakage I had was on a nightly-only project, I just noted that I needed to add a feature flag to a crate that didn't need any originally (so no real problem at all).

I just created this ticket to make sure that it was noticed and properly documented.

[alexcrichton]

Ok, thanks for the info and heads up! We'll discuss this at the next meeting and can decide what to do.

[thepowersgang]

Note - I am all for the new API, so if this change doesn't impact anyone else then probably just as good to keep it.

[alexcrichton]

The libs team discussed this regression during triage today and the decision was that this is acceptable for now. This is a clear case of fixing a bug. We'll do a crater run to analyze more crates and see what pops up, though.