`cargo metadata` exited with an error: error: could not execute process `rustc -vV` (never executed)

[Geobert]

rust-analyzer version: 0.4.1840

rustc version: 1.68.2

Since version 0.4.1840 RA can’t start:

2024-02-14T08:42:24.398537Z ERROR rust_analyzer::main_loop: FetchWorkspaceError:
rust-analyzer failed to load workspace: Failed to load the project at /home/geobert/work/project/Cargo.toml: Failed to read Cargo metadata from Cargo.toml file /home/geobert/work/project/Cargo.toml, Some(Version { major: 1, minor: 68, patch: 2 }): Failed to run `cd "/home/geobert/work/project" && "/home/geobert/.rustup/toolchains/1.68.2-x86_64-unknown-linux-gnu/bin/cargo" "metadata" "--format-version" "1" "--manifest-path" "/home/geobert/work/project/Cargo.toml" "--filter-platform" "x86_64-unknown-linux-gnu"`: `cargo metadata` exited with an error: error: could not execute process `rustc -vV` (never executed)

Caused by:
  No such file or directory (os error 2)

In VSCode on WSL2

0.4.1838 works fine

[Veykril]

Hmm, why does cargo metadata not find rustc? This is most likely caused by #16537, we now invoke cargo with the full sysroot path, but that shouldn't have an effect here now for it to execute rustc should it?

[oing9179]

rust-analyzer version: rust-analyzer version: 0.4.1840-standalone (3770f73bd 2024-02-12)
rustc version: rustc 1.76.0 (07dca489a 2024-02-04)

An issue happened to me and I believe they have something in similar, I'm on
windows and here's my error output of Rust Analyzer Language Server:

2024-02-14T09:15:58.338514Z ERROR rust_analyzer::main_loop: FetchBuildDataError:
error: failed to run `rustc` to learn about target-specific information

Caused by:
  process didn't exit successfully: `c:\Users\fuckms\.vscode-oss\extensions\rust-lang.rust-analyzer-0.4.1840-win32-x64\server\rust-analyzer.exe rustc - --crate-name ___ --print=file-names --crate-type bin --crate-type rlib --crate-type dylib --crate-type cdylib --crate-type staticlib --crate-type proc-macro --print=sysroot --print=split-debuginfo --print=crate-name --print=cfg` (exit code: 101)
  --- stderr
  program not found

I installed Rust in MSYS2 with environment variable RUSTUP_HOME and
CARGO_HOME set to /home/<Username>/.rustup and /home/<Username>/.cargo
respectively. In order to make Rust Analyzer work for my environment I have to
set rust-analyzer.server.extraEnv and 4 simliar others with 2 env vars
previously mentioned to C:\\msys64\\home\\<Username>\\.rustup and .cargo.
I suspect the issue is caused by this commit,
because on Rust Analyzer verison 0.3.1839 it worked just fine.

Edit: missing information.

[lnicola]

Child processes will inherit environment variables from their parent process by default. Environment variables explicitly set using Command::envs take precedence over inherited variables. You can disable environment variable inheritance entirely using Command::env_clear or for a single key using Command::env_remove.

Not sure why it would be a981db5, then.

[Veykril]

process didn't exit successfully: c:\Users\fuckms\.vscode-oss\extensions\rust-lang.rust-analyzer-0.4.1840-win32-x64\server\rust-analyzer.exe rustc - --crate-name ___ --print=file-names --crate-type bin --crate-type rlib --crate-type dylib --crate-type cdylib --crate-type staticlib --crate-type proc-macro --print=sysroot --print=split-debuginfo --print=crate-name --print=cfg (exit code: 101)

🤨 That looks like the rustc_wrapper causing troubles somehow

[Veykril]

a981db5 should be irrelevant here as that doesn't involve build scripts.

[Veykril]

So it has to be #16537. Instead of running cargo from CARGO_HOME etc we invoke it from the sysroot directly (so cargo no longer does any toolchain selection here) which presumably changes how additional tools are invoked? I would assume cargo would just pick the rustc besides it in the toolchain when invoked this way...