Improve inline asm error diagnostics

[Amanieu]

Previously we were just using the raw LLVM error output (with line, caret, etc) as the diagnostic message, which ends up looking rather out of place with our existing diagnostics.

The new diagnostics properly format the diagnostics and also take advantage of LLVM's per-line srcloc attribute to map an error in inline assembly directly to the relevant line of source code.

Incidentally also fixes #71639 by disabling srcloc metadata during LTO builds since we don't know what crate it might have come from. We can only resolve srclocs from the currently crate since it indexes into the source map for the current crate.

Fixes #72664
Fixes #71639

r? @petrochenkov

Old style

#![feature(llvm_asm)]

fn main() {
    unsafe {
        let _x: i32;
        llvm_asm!(
            "mov $0, $1
             invalid_instruction $0, $1
             mov $0, $1"
             : "=&r" (_x)
             : "r" (0)
             :: "intel"
        );
    }
}

error: <inline asm>:3:14: error: invalid instruction mnemonic 'invalid_instruction'
             invalid_instruction ecx, eax
             ^~~~~~~~~~~~~~~~~~~

  --> src/main.rs:6:9
   |
6  | /         llvm_asm!(
7  | |             "mov $0, $1
8  | |              invalid_instruction $0, $1
9  | |              mov $0, $1"
...  |
12 | |              :: "intel"
13 | |         );
   | |__________^

New style

#![feature(asm)]

fn main() {
    unsafe {
        asm!(
            "mov {0}, {1}
             invalid_instruction {0}, {1}
             mov {0}, {1}",
            out(reg) _,
            in(reg) 0i64,
        );
    }
}

error: invalid instruction mnemonic 'invalid_instruction'
 --> test.rs:7:14
  |
7 |              invalid_instruction {0}, {1}
  |              ^
  |
note: instantiated into assembly here
 --> <inline asm>:3:14
  |
3 |              invalid_instruction rax, rcx
  |              ^^^^^^^^^^^^^^^^^^^

[Amanieu]

I think we could get even better diagnostics if we could encode more data into the srcloc. LLVM only gives us one u32 per asm source line so we would need to store an index into a table of asm lines. I haven't looked into that yet but it may get quite hairy due to interactions with incremental compilation and codegen units. I would appreciate some mentoring there!

[mominul]

Would this fix #72664 ?

[Amanieu]

Ah yes, indeed.

[petrochenkov]

This is a great improvement.
I can't help with the FIXME, unfortunately.
@bors r+

[bors]

📌 Commit bd7db0fb1ed741f8ef0342dd69797e4117af0f04 has been approved by petrochenkov

[estebank]

It would be nice if this could point at the entire width of invalid_instruction. (I saw the FIXME, which is why this isn't the case.)

[estebank]

It would be great if we could avoid the note when it is redundant and the resulting evaluated assembly is the same as what is in the code.

[Amanieu]

@bors r=petrochenkov

[bors]

📌 Commit fc497f7 has been approved by petrochenkov