Unresolved import error for "ed25519_dalek::Keypair"

[pmuens]

First of all: Thanks for creating rust-analyzer.

I'm using the VSCode extension version 0.2.449 and I'm running into the "rust-analyzer(unresolved-import)" issue which was already reported before.

I followed the advice to enable the following settings to no avail:

{
  "rust-analyzer.cargo.loadOutDirsFromCheck": true,
  "rust-analyzer.procMacro.enable": true
}

Here's my full settings.json file (I've restarted VSCode as prompted after changing the extension settings):

{
  "editor.tabSize": 4,
  "editor.formatOnSave": true,
  "editor.formatOnPaste": true,
  "editor.formatOnType": true,
  "files.insertFinalNewline": true,
  "files.trimTrailingWhitespace": true,
  "[rust]": {
    "editor.defaultFormatter": "matklad.rust-analyzer"
  },
  "rust-analyzer.cargo.loadOutDirsFromCheck": true,
  "rust-analyzer.procMacro.enable": true
}

Here's the code snippets which causes the error to pop up:

// Cargo.toml dependency: ed25519-dalek = "1.0.1"
use ed25519_dalek::Keypair; // <--

Let me know if you need anything else and thanks in advance for looking into this 👍

[Veykril]

The problem is this line https://docs.rs/ed25519-dalek/1.0.1/src/ed25519_dalek/lib.rs.html#238 I believe. Rust analyzer checks with test enabled so the entire crate looks empty to it.

[lnicola]

CC #7225

[jonas-schievink]

I guess what we actually want is to disable test when a crate is used as a dependency, but enable it when opening the crate in the editor (unless the user disables it there, as per #7225).

No idea if/how this would make sense to the user though.

[lnicola]

I suppose there's no chance of having something like "assume --cfg test while editing code gated by #[cfg(test)]", right?

[jonas-schievink]

That would actually be my preferred solution. Not sure how feasible it is to implement though.

[flodiebold]

I think both things you wrote are basically the same 😄 I.e. we need to have each crate (or at least each leaf crate) in a test and non-test version in the crate graph. Then hopefully the "assume --cfg test while editing code gated by #[cfg(test)]" would happen automatically as long as our "selecting the crate from the cursor position" logic is smart enough?

[mieubrisse]

Adding a datapoint that I'm getting the exact same error:

// Cargo.toml: ed25519-dalek = "1.0.1"
use ed25519_dalek::Keypair;

[jplatte]

dalek_ed25519 issue: dalek-cryptography/ed25519-dalek#173

[jonas-schievink]

We should make this work by default