Change error `contains interior mutability` to `may contain interior mutability`?

[dpc]

The following error is a bit misleading:

the type SomeTrait + 'static contains interior mutability and a reference may not be safely transferrable across a catch_unwind boundary

Since SomeTrait is a trait object, it seems to me may contain (...) is more precise.

[Mark-Simulacrum]

@rust-lang/lang: Do we want to make this change? If so, E-easy and E-mentor on me. I think it makes sense to edit the language; unfortunately I don't think we can do it only when referencing a trait object.

This message needs to be updated to the new language and src/test/compile-fail/issue-17718-const-borrow.rs and src/librustc_mir/diagnostics.rs need updating as of the time of writing this.

[nikomatsakis]

may contain seems fine. We are perhaps a bit inconsistent on this -- it often happens that we make static approximations, and hence all the compiler's conclusions are sort of (implicitly) "may contain foo". Seems like it's probably better to try to be explicit about that in our phrasing, though, since it can really throw people otherwise.

[Mark-Simulacrum]

Okay, I've marked as E-easy and E-mentor; initial instructions are here. Let us know if you'd like to work on this or have any questions, we'd be happy to help.

[gaurikholkar-zz]

I'd like to take this up

[Mark-Simulacrum]

Go ahead! Let us know if you'd like any assistance.

[gaurikholkar-zz]

So we aren't doing it only while referencing a TraitObject right?

[Mark-Simulacrum]

No, since that'd require doing this inside the compiler; we can get 95% of the way there by just changing the error message. There's not a good way to check in the annotation if the Self type is a trait object.

[gaurikholkar-zz]

Here's a link to my local branch. I've made the changes as you suggested followed by a ./x.py build --stage 0 and ./x.py build --stage 1. But the compile-fail test seems to be failing. The change does not seem to get reflected somehow. Do I need to do something more?

[Mark-Simulacrum]

Could you post a build log? By the way, I think to test you should run ./x.py test --incremental src/test/compile-fail in this case and that should be good.

[eddyb]

You changed RefUnwindSafe's on_unimplemented message, which is relevant to this issue, but none of the testcases you changed actually use RefUnwindSafe (they just happen to have similar phrasing, but they're about borrowing values with UnsafeCell in them, in constants).

[gaurikholkar-zz]

@eddyb thanks! I happened to not notice that these two were indeed different. The error message regarding borrowing values appears here . That'll make the compile-fail test work. I'll write a new ui test for the RefUnwindSafe case as discussed with @nikomatsakis .

[gaurikholkar-zz]

Updated the code here
./x.py test src/test/compile-fail works
./x.py test src/test/ui works
Should I open up a PR?