Test: Ambigious bindings in same namespace with the same res

[LorrensP-2158466]

Add a test based on the discussion here and related to #145575 (comment).

This is the most reduced form I could create that passes on nightly but fails with #145108 (see #gsoc > Project: Parallel Macro Expansion @ 💬).

Also not sure about the test names.

r? @petrochenkov

[petrochenkov]

This is the most reduced form I could create that passes on nightly

I mostly wanted a version of the same issue that fails on nightly, but having both would be even better.

[LorrensP-2158466]

Added a revision that uses the version where the glob import is inserted by a macro. I have confirmed this fails when using cargo +stage1 check on my own crates, but for some reason, this is not working in ui tests. I'm probably doing something wrong, but I can't figure it out from the docs.

[LorrensP-2158466]

@rustbot ready

I'm still confused why the nightly-fail doesn't fail in the tests but fails when I extract them into their own separate crates and compile them myself.

[petrochenkov]

@rustbot ready

I'm still confused why the nightly-fail doesn't fail in the tests but fails when I extract them into their own separate crates and compile them myself.

Need to figure this out, split the test if necessary.
@rustbot author

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[petrochenkov]

Default edition may be different in ui tests (2015) and your local setup, it affects import resolution.

[LorrensP-2158466]

Ah, right, editions. I set them up correctly based on the output I get on my local setup with cargo +stage1 build -v. Both crates are compiled with edition 2018, and the pulling crate in 2024. The error is still different, with it now reporting that the macro is private instead of not accessible in the scope.

The only difference I see is that ambiguous_extern is passed as a dylib in the tests and as a .rmeta file in my local setup. I have been trying to understand how to change this, but i have not made any progress.

@rustbot ready

[petrochenkov]

The error is still different, with it now reporting that the macro is private instead of not accessible in the scope.

This version is even better, the issue can be seen more directly.

The only difference I see is that ambiguous_extern is passed as a dylib in the tests and as a .rmeta file in my local setup.

This is not important.

r=me with green CI (stderr files need to be updated) and after squashing commits.

[LorrensP-2158466]

This version is even better, the issue can be seen more directly.

Nice!

r=me with green CI (stderr files need to be updated) and after squashing commits.

You'll have to r+ this :), I don't have those privileges. @rustbot ready