Add coroutine info to v0 symbol mangling (via DefPath/DefPathData)

[sharnoff]

See #104830 (comment) for some context.

The high-level goal here is to change the v0 symbol mangling to include any relevant coroutine information for closures — i.e. if it's been desugared from async, gen, or #[coroutine] blocks/closures.

Example of the intended change

So the idea is to move from this:

// mcyrate::foo::{closure#0}
async fn foo() {}

fn bar() {
    // mycrate::bar::{closure#0}
    let _ = async {};

    // mycrate::bar::{closure#1}::{closure#0}
    let _ = async || {};
    // mycrate::bar::{closure#2}::{closure#0}
    let _ = || async {};

    // mycrate::bar::{closure#3}
    let _ = #[coroutine] || {};
}

to this:

// mcyrate::foo::{closure:async#0}
async fn foo() {}

fn bar() {
    // mycrate::bar::{closure:async#0}
    let _ = async {};

    // mycrate::bar::{closure#0}::{closure:async#0}
    let _ = async || {};
    // mycrate::bar::{closure#1}::{closure:async#0}
    let _ = || async {};

    // mycrate::bar::{closure:coroutine#0}
    let _ = #[coroutine] || {};
}

---

To do that, this PR:

1. Adds coroutine information to rustc_hir::DefKind
2. Passes that information through to rustc_hir::DefPathData
3. Exposes the coroutine information via DefPathData::get_opt_name so that it shows up as {closure:<coroutine kind>} in symbol mangling
4. Uses the new DefPathData coroutine info in DefPathDataName to have similar {closure:async} formatting in other places

Ideally it wouldn't have required threading this information everywhere, but AFAICT the only other way to do it would be via accessing HIR nodes during type printing, which IIUC won't work because we can only get HIR nodes for the local crate (?)

This PR affects a few different areas, so I'm not totally sure who to ping for review. Also more than happy to split it up into a few smaller PRs, if that'd be useful.

r? rust-lang/compiler

[rustbot]

Some changes occurred in coverage tests.

cc @Zalathar

Some changes occurred in match checking

cc @Nadrieril

Some changes occurred in compiler/rustc_codegen_ssa

cc @WaffleLapkin

Some changes occurred in compiler/rustc_sanitizers

cc @rcvalle

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

Some changes occurred in src/tools/clippy

cc @rust-lang/clippy

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

HIR ty lowering was modified

cc @fmease

[compiler-errors]

I wonder if a simpler approach here would be to make a closure_kind query and encode it cross-crate like we do for coroutine_kind.

Then in the demangler, you could instead just intercept the tcx.coroutine_kind to decorate the closure with a better name than its anonymized name?

Not sure if that's possible.

[rust-log-analyzer]

The job aarch64-gnu-llvm-19-1 failed! Check out the build log: (web) (plain enhanced) (plain)

Click to see the possible cause of the failure (guessed by this bot)

--- stderr -------------------------------
warning: function `gbc` is never used
##[warning] --> coroutine_body.rs:2:18
  |
2 |         async fn gbc() -> i32 {
  |                  ^^^
  |
  = note: `#[warn(dead_code)]` on by default


thread 'rustc' panicked at /checkout/tests/ui-fulldeps/stable-mir/check_coroutine_body.rs:43:9:
Cannot find `gbc::{closure#0}`. All local items are: [
    CrateItem(
        DefId {
            id: 0,
            name: "gbc",
        },
    ),
    CrateItem(
        DefId {
            id: 1,
            name: "gbc::{closure:async#0}",
        },
    ),
    CrateItem(
        DefId {
            id: 2,
            name: "gbc::{closure:async#0}::{closure:async#0}",
        },
    ),
    CrateItem(
        DefId {
            id: 3,

[sharnoff]

That would definitely be much better 😅

AFAICT the earliest point where coroutine info is available is when HIR nodes are constructed, but those can only be accessed by LocalDefId.

I thought it wouldn't work for that reason, but from what you said - do I understand correctly that if a query has separate_provide_extern, it'll be able to get that information, even if the DefId is from a different crate?
(I guess like what's done here?)

If so, we can probably just use the coroutine_kind in the symbol mangler already (and I can open a new, much simpler PR 😄)

[bors]

☔ The latest upstream changes (presumably #143434) made this pull request unmergeable. Please resolve the merge conflicts.