cargo check doesn't give correct diagnostics

[cod1r]

Problem

Running cargo check doesn't give the correct diagnostics when using functions that don't exist in an already built module once rust-analyzer is restarted.

Steps

Example:

fn main() {
}

Then run cargo build.

Change Rust contents to:

fn main() {
    chicken();
}

Write to disk.
(Running cargo check once file is written to disk will get it to display the correct diagnostics if rust-analyzer is not restarted)
Restart rust-analyzer.
(In vim/neovim, you can run :LspRestart. In VSCode, do ctrl+shift+p (cmd+shift+p on mac), then type "restart" and click on "rust-analyzer: restart server")

Run cargo check again.

Output will be something like:

Finished dev [unoptimized + debuginfo] target(s) in 0.00s

even though the Rust contents are invalid (unknown function is being called).

Description of cargo check is:
"The compiler will save metadata files to disk so that future runs will reuse them if the source has not been modified"

Source has been modified and incorrect diagnostics are being given therefore I believe this is a bug. :)

Possible Solution(s)

running cargo clean, use cargo build instead, write file to disk again.

Notes

No response

Version

cargo 1.73.0 (9c4383fb5 2023-08-26)
release: 1.73.0
commit-hash: 9c4383fb55986096b414d98125421ab87b5fd642
commit-date: 2023-08-26
host: aarch64-apple-darwin
libgit2: 1.6.4 (sys:0.17.2 vendored)
libcurl: 8.1.2 (sys:0.4.65+curl-8.2.1 system ssl:(SecureTransport) LibreSSL/3.3.6)
ssl: OpenSSL 1.1.1u  30 May 2023
os: Mac OS 14.0.0 [64-bit]

[cod1r]

@rustbot label: +Command-check

[cod1r]

I'm not sure whether this issue belongs in the rust-analyzer repo or the cargo repo because I don't know the relationship between the two. I have been told that rust-analyzer uses cargo output. So I don't know how restarting rust-analyzer would make cargo check display the wrong information.

[weihanglo]

I reproduced it in neovim, and observed some oddities. Let me write it down:

• When you first open the editor and rust-analzyer fire cargo check, you will get a fingerprint file with 16 hex digits, say, target/debug/.fingerprint/mybin-3e3731215339b3df/bin-mybin.
• After you input some trash in your editor, rust-analyzer will fire another cargo check and the fingerprint now is nuked as expected (I believe so, see this code).
• After you restart LSP server (via :LspRestart), the 16-hexdigit fingerprint is re-created mysteriously, which shouldn't.

I didn't dig deep enough, but I believe there might be some bug in Cargo, if what rust analyzer did is just invoking cargo check with valid args.

I'll recommend also opening an issue in rust-lang/rust-analyzer and see if they have insights.