[EXPERIMENT] DelimSpan::from_single() experiment and UI tests update.

[Keith-Cancel]

Experimental changes to DelimSpan::from_single() to have potential better open and close values.

Here is the tests output updated as you asked as a separate PR as asked here: #149052 (comment)

r? @petrochenkov

[tgross35]

Could you make the PR description more meaningful? We shouldn't have to refer to a separate PR to know what this does, and it definitely does more than just update tests.

Please also make sure to squash before merge.

[Keith-Cancel]

Could you make the PR description more meaningful? We shouldn't have to refer to a separate PR to know what this does, and it definitely does more than just update tests.

Please also make sure to squash before merge.

@tgross35 I updated the title of the PR, @petrochenkov wanted a separate PR for these experiments so that's what I did. I was not sure how much detail or what should be specified in such a PR. So my bad, if there is anything else please let me know.

[rustbot]

Some changes occurred in src/tools/clippy

cc @rust-lang/clippy

[petrochenkov]

I'm not sure why the logic above is so overcomplicated.
let first = src[0] and let last = src[src.len() - 1].
open is sp.with_lo(sp.lo() + BytePos(1)) if the source map check passes, close is sp.with_hi(sp.hi() - BytePos(1)) .

[petrochenkov]

I would say a lot the differences on those tests that changed are caused by using DelimSpan::open or Delimnspan::close instead of DelimSpand::entire() somewhere when generating the diagnostic output.

This needs to be confirmed, could you check what happens at least in the case of tests/ui/did_you_mean/bad-assoc-expr.rs? Because in that case it's a clear regression and incorrect behavior.

They also all seem to involve macro metavariables.

Of course, Delimiter::Invisible means a metavariable.
That's why I was skeptical about using shrink_to_lo/shrink_to_hi instead of the whole span in practice.