Don't suggest a semicolon when one already exists

[joculatrix]

In cases where it can be confirmed that an expression with borrow conflicts is already followed by an expression, the compiler will no longer suggest adding a second semicolon.

This is my first time contributing to the compiler; I eagerly accept any feedback on improvements or how this could be done differently.

closes #133941

[rustbot]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @lcnr (or someone else) some time within the next two weeks.

Please see the contribution instructions for more information. Namely, in order to ensure the minimum review times lag, PR authors and assigned reviewers should ensure that the review label (S-waiting-on-review and S-waiting-on-author) stays updated, invoking these commands when appropriate:

• @rustbot author: the review is finished, PR author should check the comments and take action accordingly
• @rustbot review: the author is ready for a review, this PR will be queued again in the reviewer's queue

[lcnr]

Hi 👋 please add a regression test to an appropriate subfolder of tests/ui

[lcnr]

we probably want to change BlockTailInfo::tail_result_is_ignored instead. Using spans for such checks does not work in case the error involves macro calls. The current check may also have issues if something like tail_expr ; or tail_expr/* */; or tail_expr<new-line>; is used

I could imagine that we should either change tail_result_is_ignored to an enum tracking why it's ignored or alternatively change LocalInfo::BlockTailTemp to track whether the tail is an expr or a statement

[joculatrix]

That makes sense! Thanks for the suggestion - I knew someone who knew more than me would steer me in a better direction. I'll look into that. 👍

[joculatrix]

@rustbot author

[joculatrix]

@lcnr Sorry if you're not a good person to ask -- as far as I can tell, information about whether or not a statement/block ends with a semicolon is lost when hir::StmtKind::Expr and hir::StmtKind::Semi both get lowered to the same representation in the THIR, which then doesn't pass it on to the MIR. The error message in question is caught during borrow-checking, which uses the MIR. I just want to ask before doing it because I'm a new contributor and don't want to go beyond the scope of this issue/PR: is it appropriate to go back as far as editing how the HIR is lowered to THIR and how that then is lowered to MIR in order to preserve whether statements end with semicolons? Or, if not, is there some other way of getting this information that I'm missing?

[lcnr]

: is it appropriate to go back as far as editing how the HIR is lowered to THIR and how that then is lowered to MIR in order to preserve whether statements end with semicolons?

If we currently end up erasing that information, making sure it's tracked somewhere during lowering is appropriate

[joculatrix]

I've pivoted towards preserving that info in an enum as suggested. The current commit doesn't work yet and breaks a lot of tests, so it's still a work in progress, but I think I'm close (and wanted to push my work so far just so it's clear I'm working on it).