…ly verbose

Refactor `powerpc64` call ABI handling

As the [specification](https://openpowerfoundation.org/specifications/64bitelfabi/) for the ELFv2 ABI states that returned aggregates are returned like arguments as long as they are at most two doublewords, I've merged the `classify_arg` and `classify_ret` functions to reduce code duplication. The only functional change is to fix rust-lang#128579: the `classify_ret` function was incorrectly handling aggregates where `bits > 64 && bits < 128`. I've used the aggregate handling implementation from `classify_arg` which doesn't have this issue.

``@awilfox`` could you test this on `powerpc64-unknown-linux-musl`? I'm only able to cross-test on `powerpc64-unknown-linux-gnu` and `powerpc64le-unknown-linux-gnu` locally at the moment, and as a tier 3 target `powerpc64-unknown-linux-musl` has zero CI coverage.

Fixes: rust-lang#128579

…k-Simulacrum

Add windows-targets crate to std's sysroot

With this PR, when backtrace is used as a crate from crates.io it will (once updated) use the real [windows-targets](https://crates.io/crates/windows-targets) crate. But when used from std it'll use std's replacement version.

This allows sharing our customized `windows_tagets::link!` macro between std proper and the backtrace crate when used as part of std, ensuring a consistent linking story. This will be especially important once we move to using [`raw-dylib`](https://doc.rust-lang.org/reference/items/external-blocks.html#dylib-versus-raw-dylib) by default.

…iler-errors

Tidy up `dump-ice-to-disk` and make assertion failures dump ICE messages

For the future traveler: **if you did a `git blame` and found this PR that last modified `dump-ice-to-disk` because the test failed in a completely unrelated PR, then I'm afraid our ICE dump may have regressed or somehow behaves differently on `i686-mingw`.**

A bit of clean up to the `dump-ice-to-disk` test.

- Fixes/updates the top-level comment.
- Add a FIXME pointing to rust-lang#128911 for flakiness.
- Instead of trying to manually cleanup `rustc-ice*.txt` dumps, run each test instance in a separate temp directory.
- Explicitly make `RUSTC_ICE` unavailable in one of the `-Zmetrics-dir` test case to not have interference from environment.
- Make assertion failures (on ICE dump line count mismatch) extremely verbose to help debug why this test is flakey in CI (rust-lang#128911).

Contains a fix by `@saethlin` in rust-lang#128909, should wait until that is merged then rebase on top.

try-job: aarch64-gnu
try-job: aarch64-apple
try-job: x86_64-msvc
try-job: i686-mingw
try-job: test-various

…s, r=compiler-errors

Fix codegen-units tests that were disabled 8 years ago

I don't know if any of these tests still have value. They were disabled by rust-lang#33890, and we've survived without them for a while. But considering how small this test suite is, maybe it's worth having them.

I also had to add some normalization to the codegen-units tests output. I think the fact that I had to add some underscores how poor our test coverage is.

Fix warnings in rmake tests on `x86_64-unknown-linux-gnu`

r? ``@jieyouxu``

This PR fixes some warnings I saw in rmake tests. I didn't deny more warnings in this PR until ``@jieyouxu`` gives their opinion, but maybe we should actually deny all warnings in `rmake.rs` files?

I've also only looked at non-ignored tests on `x86_64-unknown-linux-gnu`, and denying warnings would require a try build for all targets 😓.