Clippy gets rebuilt twice on every change, and rustdoc also gets rebuilt

[RalfJung]

Something seems wrong with the rebuild tracking for clippy in x.py, making working on clippy in the rustc repo (e.g. when fixing breakage from a rustc change) quite slow due to the multi-minute edit-compile cycle:

./x.py test clippy --stage 2 --bless -- local
# change something in clippy_lints
./x.py test clippy --keep-stage 0 --stage 2 --bless -- local

The last command does

Building stage0 tool lld-wrapper (x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.15s
Building stage1 library artifacts (x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.06s
Building compiler artifacts (stage1 -> stage2, x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.28s
Creating a sysroot for stage2 compiler (use `rustup toolchain link 'name' build/host/stage2`)
Building stage1 tool lld-wrapper (x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.14s
Uplifting library (stage1 -> stage2)
Uplifting rustc (stage1 -> stage3)
Building tool clippy-driver (stage2 -> stage3, x86_64-unknown-linux-gnu)
   Compiling clippy_lints v0.1.83 (/home/r/src/rust/rustc/src/tools/clippy/clippy_lints)
   Compiling clippy v0.1.83 (/home/r/src/rust/rustc/src/tools/clippy)
    Finished `release` profile [optimized + debuginfo] target(s) in 1m 16s
Building tool rustdoc (stage1 -> stage2, x86_64-unknown-linux-gnu)
   Compiling rustdoc v0.0.0 (/home/r/src/rust/rustc/src/librustdoc)
   Compiling rustdoc-tool v0.0.0 (/home/r/src/rust/rustc/src/tools/rustdoc)
    Finished `release` profile [optimized + debuginfo] target(s) in 54.20s
Testing clippy (stage2 -> stage3, x86_64-unknown-linux-gnu)
   Compiling clippy_config v0.1.83 (/home/r/src/rust/rustc/src/tools/clippy/clippy_config)
   Compiling clippy_utils v0.1.83 (/home/r/src/rust/rustc/src/tools/clippy/clippy_utils)
   Compiling clippy_lints v0.1.83 (/home/r/src/rust/rustc/src/tools/clippy/clippy_lints)
   Compiling clippy v0.1.83 (/home/r/src/rust/rustc/src/tools/clippy)
    Finished `release` profile [optimized + debuginfo] target(s) in 1m 24s
     Running unittests src/driver.rs (build/x86_64-unknown-linux-gnu/stage2-tools/x86_64-unknown-linux-gnu/release/deps/clippy_driver-11f2fd5bded4b22e)

That's a lot more Compiling ... lines than I would expect, and the entire thing takes so long (>4 minutes, which is longer than it takes my machine to build all of rustc!) to make iteration and testing painful.

The most surprising part is that rustdoc gets rebuilt?!? But building clippy twice also seems like something went wrong somewhere.

[jieyouxu]

I bet 10 bors credits this is differing RUSTFLAGS again.

[onur-ozkan]

Worth to solve this problem properly along with a test coverage.. I think, all rustflags should be handled from a single place and every conditional rustflags should be propagated to bootstrap shims using environment variable (something like IMPLICIT_RUSTFLAGS) so they can't invalidate the build cache. Bootstrap should never set any conditional rustflag explicitly on cargo.

[RalfJung]

I bet 10 bors credits this is differing RUSTFLAGS again.

Having fixed some of those cases affecting Miri in the past -- that's my bet, too.