"./x.py run miri --stage 1" no longer works

[RalfJung]

This used to work fine at some point, but now leads to some strange errors:

$ ./x.py run miri --stage 1 -- src/tools/miri/tests/pass/hello.rs
Building bootstrap
    Finished `dev` profile [unoptimized] target(s) in 0.02s
Building stage0 library artifacts (x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.07s
Building compiler artifacts (stage0 -> stage1, x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.11s
Creating a sysroot for stage1 compiler (use `rustup toolchain link 'name' build/host/stage1`)
Building stage0 tool lld-wrapper (x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.06s
Building tool miri (stage0 -> stage1, x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.07s
Building tool cargo-miri (stage0 -> stage1, x86_64-unknown-linux-gnu)
    Finished `release` profile [optimized + debuginfo] target(s) in 0.07s
Building stage1 miri sysroot (x86_64-unknown-linux-gnu)
A sysroot for Miri is already available in `/home/r/src/rust/rustc.3/build/x86_64-unknown-linux-gnu/miri-sysroot`.
    Finished `release` profile [optimized + debuginfo] target(s) in 0.07s
     Running `build/x86_64-unknown-linux-gnu/stage0-tools/x86_64-unknown-linux-gnu/release/miri --target x86_64-unknown-linux-gnu src/tools/miri/tests/pass/hello.rs`
error[E0786]: found invalid metadata files for crate `std`
  |
  = note: corrupt metadata encountered in /home/r/src/rust/rustc.3/build/x86_64-unknown-linux-gnu/stage0-sysroot/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd-e9d12b2c234bc463.rlib
  = note: corrupt metadata encountered in /home/r/src/rust/rustc.3/build/x86_64-unknown-linux-gnu/stage0-sysroot/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd-e9d12b2c234bc463.so
  = note: corrupt metadata encountered in /home/r/src/rust/rustc.3/build/x86_64-unknown-linux-gnu/stage0-sysroot/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd_detect-ec2a734cf29b8414.rlib

error: cannot find macro `eprintln` in this scope
 --> src/tools/miri/tests/pass/hello.rs:3:5
  |
3 |     eprintln!("Hello, error!");
  |     ^^^^^^^^

error: cannot find macro `println` in this scope
 --> src/tools/miri/tests/pass/hello.rs:2:5
  |
2 |     println!("Hello, world!");
  |     ^^^^^^^

error: requires `sized` lang_item

error: aborting due to 4 previous errors

For more information about this error, try `rustc --explain E0786`.
Build completed unsuccessfully in 0:00:01

[Noratrieb]

somehow Miri searches for its sysroot in stage0-sysroot instead of its own built sysroot?

[Noratrieb]

When executing Miri:

[src/tools/miri/src/bin/miri.rs:359:5] std::env::var("MIRI_SYSROOT") = Ok(
    "/home/nora/projects/rust/build/x86_64-unknown-linux-gnu/miri-sysroot",
)
[src/tools/miri/src/bin/miri.rs:360:5] std::env::args().collect::<Vec<_>>() = [
    "build/x86_64-unknown-linux-gnu/stage0-tools/x86_64-unknown-linux-gnu/release/miri",
    "--target",
    "x86_64-unknown-linux-gnu",
    "uwu2.rs",
]

I thought the sysroot was passed to the process with --sysroot, but I cannot find such an argument... how exactly does it find stage0-sysroot from this... I don't really understand how that works. But this may help.

[Noratrieb]

I see, it gets the sysroot because librustc_driver.so is in stage0-sysroot. Slightly strange, that sysroot is not really one it can use. Why is it linked from there? But either way, the sysroot we want to use for Miri here is an entirely different one that needs to be override manually. How did that work previously?

[Noratrieb]

oh lol, this was almost certainly broken by rust-lang/miri#3411, in which you even specifically mentioned x.py run, looks like you forgot about it :D

[RalfJung]

*oops* thanks for fixing!