Rust 1.20 regression, for_each 0.2.0, conflicts with unstable iterator_for_each feature

[jdm]

   Compiling foreach v0.2.0 (file:///Users/jdm/src/iter_foreach)
error: use of unstable library feature 'iterator_for_each' (see issue #42986)
  --> tests/basic.rs:10:10
   |
10 |     iter.for_each(|item, iter| {
   |          ^^^^^^^^
   |
   = help: add #![feature(iterator_for_each)] to the crate attributes to enable

error[E0593]: closure takes 2 arguments but 1 argument is required
  --> tests/basic.rs:10:10
   |
10 |       iter.for_each(|item, iter| {
   |  __________^^^^^^^^_-
   | |          |
   | |          expected closure that takes 1 argument
11 | |                       assert_eq!(item, i * 2);
12 | |                       i += 1;
13 | |                       let _ = iter.next();
14 | |                   });
   | |___________________- takes 2 arguments

error: aborting due to 2 previous errors

error: Could not compile `foreach`.
warning: build failed, waiting for other jobs to finish...
error: build failed

cc @jaje

It appears that a new unstable feature was implemented that conflicts with the for_each crate. rustc interprets the use of Iterator::for_each as attempting to use the unstable feature instead of the trait method defined by the for_each crate.

[oli-obk]

I think this is expected breakage, since you could use UFC syntax to call the method you mean. Otherwise we could never add new methods to any type or trait.

[cuviper]

I'm surprised that it just assumes the code wanted Iterator::for_each, rather than reporting an ambiguity between that and ForEach::for_each. I thought only inherent methods took precedence like that. Is it because Iterator is in the prelude?

[sfackler]

If iter is a generic T: Iterator or an &mut Iterator trait object, Iterator::for_each is treated somewhat inherently I believe. Seems like something's busted in the check if iter is a concrete type though.

[cuviper]

In that test iter is a concrete type, a range.

[alexcrichton]

The libs team decided to hold off on evaluating what to do here until we have a crater report to see what the brekage is.

[alexcrichton]

A crater report for beta happened, and I don't believe we saw any other regressions related to this, so I'm going to close this now.