bootstrap: ./x.py test mir-opt dirties build cache for standard library

[RalfJung]

To reproduce:

./x.py test ui -- mir
# The next command should not have to build a new standard library,
# since it was already built. And yet, a rebuild happens.
./x.py test mir-opt
# Another rebuild happens here
./x.py test ui -- mir

Cc @saethlin this may be related to the special hacks the mir-opt test suite uses for standard libraries for other targets.

[onur-ozkan]

The cache gets broken due to using different rustflags in the following code:

rust/src/bootstrap/src/core/build_steps/compile.rs

Lines 239 to 265 in 05e692a

	let mut cargo = if self.is_for_mir_opt_tests {
	let mut cargo = builder::Cargo::new_for_mir_opt_tests(
	builder,
	compiler,
	Mode::Std,
	SourceType::InTree,
	target,
	Kind::Check,
	);
	cargo.rustflag("-Zalways-encode-mir");
	cargo.arg("--manifest-path").arg(builder.src.join("library/sysroot/Cargo.toml"));
	cargo
	} else {
	let mut cargo = builder::Cargo::new(
	builder,
	compiler,
	Mode::Std,
	SourceType::InTree,
	target,
	Kind::Build,
	);
	std_cargo(builder, target, compiler.stage, &mut cargo);
	for krate in &*self.crates {
	cargo.arg("-p").arg(krate);
	}
	cargo
	};

. In the first block bootstrap sets -Zalways-encode-mir directly but in the second block it uses std_cargo to handle rustflags (and other cargo-specific environment variables).

[jieyouxu]

Triage: I believe this was later superseded by #131437.

[xry111]

Triage: I believe this was later superseded by #131437.

#131437 was closed months ago but IIRC this still happens with rustc-1.85.0.

[jieyouxu]

I'm not observing the rebuild locally, please open a new issue with repro steps if you can still repro on latest nightly.