Pass FnAbi to find_mir_or_eval_fn

[tiif]

rust-lang/miri#4013 needs information from FnAbi, hence it is passed to find_mir_or_eval_fn.

r? @RalfJung

[rustbot]

Some changes occurred to the CTFE machinery

cc @rust-lang/wg-const-eval

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

The Miri subtree was changed

cc @rust-lang/miri

[RalfJung]

@rustbot author

[tiif]

@RalfJung There are some refactorings needed to remove ExternAbi from find_mir_or_eval_fn. Specifically, all functions that call this.check_shim will need to be updated. Could I confirm if this direction is what we are going for?

[RalfJung]

Specifically, all functions that call this.check_shim will need to be updated.

Yeah, they should pass fn_abi to check_shim, and all the ExternAbi::C { unwind: false } become abi::Conv::C.

[bors]

☔ The latest upstream changes (presumably #134052) made this pull request unmergeable. Please resolve the merge conflicts.

[tiif]

Interesting... for the same test, running on Linux will pass, but running on Windows will fail because of incompatible ABI.

For future reference:

• ./x.py test miri --test-args tests/pass/align.rs --target x86_64-pc-windows-gnu failed with calling a function with ABI X86Stdcall using caller ABI C
• ./x.py test miri --test-args tests/pass/align.rs --target x86_64-unknown-linux-gnu pass.

This means that for the exact same function, whether we should use Conv::C or Conv::X86Stdcall depends on the target we are running? Or maybe I just did something bad somewhere...

EDIT: Ok nevermind, it's just I used Conv::X86Stdcall when it supposed to be Conv::C, even for the same tests, it hits different path for Windows and Linux.

[RalfJung]

This means that for the exact same function, whether we should use Conv::C or Conv::X86Stdcall depends on the target we are running? Or maybe I just did something bad somewhere...

Yes that makes sense, Windows has some special rules here.

[tiif]

I might continue to fix some cosmetic issues such as stray new lines, but this PR is ready for review

@rustbot ready

[RalfJung]

Looks pretty good. :) Just some minor comments.

[RalfJung]

Just use debug formatting for the error for now. We do the same in interpret/call.rs already.

The proper fix here would be a impl Display for Conv somewhere near this. If you want you can do that in a future PR. :)

[tiif]

Sure, I can open one after this.

[RalfJung]

@rustbot author

[bors]

☔ The latest upstream changes (presumably #134381) made this pull request unmergeable. Please resolve the merge conflicts.

[tiif]

I will need some time to figure out how to resolve the merge conflict, and will request for r+ again after resolving it.

[tiif]

Hmm this is weird, the import is valid and I didn't change anything from attr, but got this error:

 --> compiler/rustc_hir/src/hir.rs:5:5
  |
5 | use rustc_ast::attr::AttributeExt;
  |     ^^^^^^^^^^^^^^^^^------------
  |     |                |
  |     |                help: a similar name exists in the module: `Attribute`
  |     no `AttributeExt` in `attr`

error[E0432]: unresolved import `rustc_ast::UnsafeBinderCastKind`
  --> compiler/rustc_hir/src/hir.rs:14:57
   |
14 |     ImplPolarity, IsAuto, Movability, Mutability, UnOp, UnsafeBinderCastKind,
   |                                                         ^^^^^^^^^^^^^^^^^^^^ no `UnsafeBinderCastKind` in the root

error[E0432]: unresolved import `rustc_ast::attr::AttributeExt`
  --> compiler/rustc_hir/src/lang_items.rs:10:5
   |
10 | use rustc_ast::attr::AttributeExt;
   |     ^^^^^^^^^^^^^^^^^------------
   |     |                |
   |     |                help: a similar name exists in the module: `Attribute`
   |     no `AttributeExt` in `attr`

For more information about this error, try `rustc --explain E0432`.

@oli-obk since you reviewed #134381, do you know what's going on here and how to resolve this?

[oli-obk]

That's weird. Sounds like a cache issue where it doesn't think a crate has changed when it acrually has. Worst case try x clean

[tiif]

x clean works, thanks!

[bors]

☔ The latest upstream changes (presumably #134243) made this pull request unmergeable. Please resolve the merge conflicts.

[RalfJung]

@bors r+

[oli-obk]

@bors ping

[bors]

😪 I'm awake I'm awake

[oli-obk]

@bors r=RalfJung

[bors]

📌 Commit fd8b983 has been approved by RalfJung

It is now in the queue for this repository.