Remove HasLocalDecls.

[nnethercote]

It's a trait that lets you pass any of Body, LocalDecls, or IndexVec<Local, LocalDecl> to various functions. A minor convenience, and a minor obfuscation, that not actually needed.

This commit removes the trait and changes those functions to receive &LocalDecls instead of &D where D: HasLocalDecls. This makes lots of call sites slightly more verbose, but I think it's worth it for the overall simplicity.

r? @scottmcm

[rustbot]

Some changes occurred in src/tools/clippy

cc @rust-lang/clippy

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

Some changes occurred in compiler/rustc_codegen_cranelift

cc @bjorn3

[compiler-errors]

This makes lots of call sites slightly more verbose, but I think it's worth it for the overall simplicity.

I disagree that this is worthwhile, and I think this existing abstraction improves code readability significantly in dense MIR like MIR typeck.

If you think this is making folks confused about what to pass in the argument, then I'd suggest leaving a doc comment on HasLocalDecls saying what implements it.

[scottmcm]

I'm happy to review this, but since I'm not actually on compiler contributors I think I'm poorly positioned to make a policy decision here on whether this is desirable, so I'll leave this to the team to talk about approach first.

[nnethercote]

My original motivation was I had a chang (not in this PR) where a call to ty(body) didn't borrowck because another field of body was also being borrowed. Changing it to ty(&body.local_decls) fixed that borrowck error. Then I felt that HasLocalDecls was unnecessary cleverness, that often the simple obvious way to do things is the best.

@scottmcm: I asked you for review because you modified some of this code in #109819, though I guess that was part of a larger change.

[wesleywiser]

We discussed in the weekly triage meeting but did not come to a consensus that the changes in this PR were more understandable than the current code. There seemed to be strong support for documenting how to deal with the borrowck error @nnethercote originally ran into in the doc comments for HasLocalDecls.

[scottmcm]

@nnethercote TBH I'm generally personally opposed to things like the "into trick", and thus have no issues with this change, I just feel like deciding that for rustc is outside my purview.

I only added the new HasLocalDecls in that PR because taking impl HasLocalDecls disables deref coercion, so the added impl is basically doing that coercion manually -- and losing coercions like this is one of the reasons that I tend to dislike passing parameters this way.

It does seem like perhaps not all the implementations make it into https://doc.rust-lang.org/nightly/nightly-rustc/rustc_middle/mir/trait.HasLocalDecls.html#implementors today? That seems suboptimal... No, I don't know what I'm thinking -- they're all there.