Reproduce bug in clippy-driver

[daivinhtran]

Steps to reproduce #2418

1. https://github.com/dmeister/rules_rules_clippy_repo and replace http_archive(name = "rules_rust") to local_repository pointing to a local repo of rules_rust
2. In rules_rust repo, apply the patch https://github.com/bazelbuild/rules_rust/pull/2444.diff
3. Flip the counter back and forth between 22 and 23 to reproduce the bug

I believe when there are too many arguments to clippy-driver, it breaks.

[daivinhtran]

Intererestingly //crate_universe:cargo_bazel_bin has similar setup but clippy-driver doesn't fail on it.

[UebelAndre]

So the complaint about duplicate --sysroot is a red herring?

[daivinhtran]

So the complaint about duplicate --sysroot is a red herring?

I believe so. Have you seen #2447?

process_wrapper parses (and processes?) the args and pass "@bazel-out/k8-fastbuild/bin/hello_world/hello_world.clippy.ok-0.params.expanded", "-Lnative=/tmp/bazel-working-directory/_main/bazel-out/k8-fastbuild/bin/external/crate_index__ring-0.16.20/ring_build_script.out_dir", "-Lnative=/tmp/bazel-working-directory/_main/bazel-out/k8-fastbuild/bin/external/crate_index__ring-0.17.7/ring_build_script.out_dir" to clippy-driver.

clippy-driver then sets --syroot based on SYSROOT.

I believe, at runtime within the execution of the process_wrapper, clippy-driver sees another --sysroot in the .expanded file and fails. That's why either removing "--sysroot" (which causes it to be removed from .expanded) or removing SYSROOT (which causes --sysroot not to be created by clippy-driver before calling rustc) fixed the error; but it's not the true fix. The true fix should be fixing process_wrapper to not wrap --sysroot into @bazel-out/k8-fastbuild/bin/hello_world/hello_world.clippy.ok-0.params.expanded or make @bazel-out/k8-fastbuild/bin/hello_world/hello_world.clippy.ok-0.params.expanded more deterministic.