Add accurate targeting for imperfect derives diagnostic in E0277

[raushan728]

Resolves #157117
Follow-up #158764

This PR implements accurate targeting for the E0277 "imperfect derives" suggestion so it only triggers when all fields of the ADT satisfy the trait independently of the failing generic parameter. False positives (like struct S<T>(T)) are now successfully filtered out, while true cases (like PhantomData<T>) still correctly emit the help note.

r? @estebank
cc @mejrs @khyperia

[khyperia]

neat! I haven't looked through the PR at all, just wanted to suggest having a test with associated types as well, not just PhantomData. You can find an example of what that looks like in my comment here - feel free to rework it to be cleaner or whatever, but also copypasting it directly would work as well, up to you. #158764 (comment)

Perhaps also Rc with Clone, and also some coinductive shenanigans too (the List type from Niko's blog post we were talking about in the previous PR)

[raushan728]

Thanks! Will update shortly.

[raushan728]

just wanted to suggest having a test with associated types as well, not just PhantomData.

Done! please take a look.

[raushan728]

Hi @mejrs @khyperia @estebank this PR has been open for a while and I haven't received any review yet. If you could help me out a bit, we can get this landed sooner 🙂

[estebank]

My apologies, the logic is a bit subtle and that makes it harder to follow. Would you mind adding more comments explaining what each section is accomplishing?

I believe the changes to be directionally correct, but suspect there might be helper methods to construct the obligation with fewer moving parts.

I am currently traveling, so my response rate is lower than it otherwise would be.

View changes since this review

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[raushan728]

My apologies, the logic is a bit subtle and that makes it harder to follow. Would you mind adding more comments explaining what each section is accomplishing?

Added

I believe the changes to be directionally correct, but suspect there might be helper methods to construct the obligation with fewer moving parts.

I looked but didn't find an existing helper for this exact "swap self_ty, keep the rest of the trait args" pattern, so I kept the explicit construction. Open to suggestions if you know of something more idiomatic.

I am currently traveling, so my response rate is lower than it otherwise would be.

No worries, thanks for taking the time to review!