Why rust-analyzer doesn't work in [test] function?

[Wyvern]

#[test]
fn test_fn(){
... //rust-analyzer doesn't work here! NO error prompt until you really compile file.
}

Is it a bug or a config/setting which should properly toggled?

[lnicola]

Does it work with #[test]?

[Wyvern]

Does it work with #[test]?

It's a typo, fixed it.

[lnicola]

It should work, can you make a sample test case?

[Wyvern]

Ok. Let's put a non-existed function call in test, there's NO error prompt until you run test.

[lnicola]

It's marked as unresolved:

rust-analyzer doesn't (yet) show errors for these because there would be too many false positives. But you can see them using something like:

{
    "editor.semanticTokenColorCustomizations": {
        "rules": {
            "unresolvedReference": "#ff0000",
        },
    }
}

[flodiebold]

I think @Wyvern might be missing rustc errors from check-on-save? @Wyvern do you have any non-default settings? In particular rust-analyzer.checkOnSave.allTargets or rust-analyzer.checkOnSave.target?

[Wyvern]

"checkOnSave" enabled. Furthermore, I also enabled auto-save editing; problem stays.

By the way, rust-analyzer works normally in non-test scope. So, it's definitely the problem specifically to #[test] function.

[flodiebold]

Do you have rust-analyzer.checkOnSave.allTargets or rust-analyzer.checkOnSave.target set?

[mrpink76]

A related fix for me was to add "--all-targets" to my tasks.json file in vscode. That way both tests and normal execution was compiled:

tasks.json

{
	"version": "2.0.0",
	"tasks": [
		{
			"type": "cargo",
			"command": "build",
			"args": [
				"--all-targets"
			],
			"problemMatcher": [
				"$rustc"
			],
			"group": "build",
			"label": "rust: cargo build",
			"presentation": {
				"clear": true
			}
		}
	]
}

[Wyvern]

I found that this problem not only affect #[test] code but also [[example]] code file, still not sure how to fix it.

[nikkolasg]

@Wyvern I confirm that --all-targets need to be enabled. It should probably be enabled by default, there's no reason to avoid showing errors on test. For reference, my setup on vscode is

"rust-analyzer.checkOnSave.overrideCommand": [
        "cargo",
        "+nightly",
        "clippy",
        "--all",
        "--all-targets",
        "--all-features",
        "--message-format=json",
        "--",
        "-D",
        "clippy::all"
    ],

[pitaj]

I was able to resolve this similar to @nikkolasg but I had to enable checking tests, examples, and benches explicitly as well:

    "rust-analyzer.checkOnSave.overrideCommand": [
        "cargo",
        "+nightly",
        "clippy",
        "--workspace",
        "--all-targets",
        "--all-features",
        "--tests",
        "--examples",
        "--benches",
        "--message-format=json",
    ],

[nobane]

The setting is now "rust-analyzer.check.overrideCommand"

[davidbarsky]

Looks like this issue can be resolved.

[TomzBench]

i have this issue with my repo also. It seems some projects, the example files will emit lint errors, while other projects will not emit lint errors.

here is more details that I originally raised in my IDE, however, it seemed RA related.

I tried changing the override command, and also explicitly adding examples my Cargo.toml file. No luck however.

[shripadk]

Had the same issue. The solution is to enable experimental diagnostics mode. My neovim specific lsp config:

-- snip
["rust_analyzer"] = function()
  require 'lspconfig'.rust_analyzer.setup {
    settings = {
      ["rust-analyzer"] = {
        diagnostics = {
          disabled = "unlinked-file",
          experimental = {
            enable = true,
          },
        },
      },
    },
  }
end,
-- snip

and for VS Code users, I think this should suffice:

"rust-analyzer.diagnostics.experimental.enable": true

[TomzBench]

Had the same issue. The solution is to enable experimental diagnostics mode. My neovim specific lsp config:

-- snip
["rust_analyzer"] = function()
  require 'lspconfig'.rust_analyzer.setup {
    settings = {
      ["rust-analyzer"] = {
        diagnostics = {
          disabled = "unlinked-file",
          experimental = {
            enable = true,
          },
        },
      },
    },
  }
end,
-- snip

and for VS Code users, I think this should suffice:

"rust-analyzer.diagnostics.experimental.enable": true

Thank you - this was bothering me so much. I'm not sure why some projects need this and others do not.