Lint emitted in UI test but not when running against a separate package

[mattvmsft]

I am having some trouble getting a lint emitted when running cargo dylint --all against a package. I have a small test lint as follows:

dylint_linting::declare_early_lint! {
    pub LINT_EXAMPLE,
    Warn,
    "API banned_api is not allowed"
}

impl EarlyLintPass for LintExample {
    fn check_fn(&mut self, cx: &rustc_lint::EarlyContext<'_>, fn_kind: rustc_ast::visit::FnKind<'_>, span: rustc_span::Span,_:rustc_ast::NodeId) {
                
        if_chain! {
            if let FnKind::Fn(_, ident, ..) = fn_kind;
            if ident.name.as_str() == "banned_api";
            then {
                span_lint_and_sugg(cx, LINT_EXAMPLE, span, "This API is banned", "it can be replaced with", "allowed_api()".into(), Applicability::MachineApplicable)
            }
        }
        
    }
}

When running the UI test for this lint on this main.rs file in my ui/ directory:

fn banned_api() {
    
}

fn main() {
    banned_api()
}

the generated stderr file contains the lint I expect:

  --> $DIR/main.rs:1:1
   |
LL | / fn banned_api() {
LL | |     
LL | | }
   | |_^ help: it can be replaced with: `allowed_api()`
   |
   = note: `-D lint-example` implied by `-D warnings`

error: aborting due to previous error

(although as a side issue, I am unable to get the UI test to pass despite copying the generated .stderr into the main.stderr in my ui/ directory. The output does not have any + or - to denote differences between the .stderr files. Not sure if that is related.)

In a separate directory, I created a new Rust binary project where the main.rs file has the same contents as that of my ui test main.rs file, but when I run cargo dylint --all I am simply getting:

Checking with toolchain `nightly-2022-09-29-x86_64-pc-windows-msvc`
    Checking example_project v0.1.0 (<sample_proj_path>)
    Finished dev [unoptimized + debuginfo] target(s) in 0.20s

without any additional output.

cargo dylint list appears to find the correct lint library:

 ~#@❯ cargo dylint list
lint_example  nightly-2022-09-29-x86_64-pc-windows-msvc  <path_to_lint_dir>

Operating System: Windows 11

[Samuel Moelius]

(although as a side issue, I am unable to get the UI test to pass despite copying the generated .stderr into the main.stderr in my ui/ directory. The output does not have any + or - to denote differences between the .stderr files. Not sure if that is related.)

Could you try changing the types of newlines that the .stderr file uses (i.e., LF vs. CRLF) and see if that makes a difference?

In a separate directory, I created a new Rust binary project where the main.rs file has the same contents as that of my ui test main.rs file, but when I run cargo dylint --all I am simply getting...

By chance, have you tried cargo clean and then cargo dylint --all?

Cargo caches linting results, and I am wondering if that could be the problem.

[Samuel Moelius]

Also, aside: are you tied to using an early lint?

I generally recommend writing late lints, unless you're absolutely sure your lint will never need typing and/or name resolution information.

[mattvmsft]

Both of your suggestions resolved my issues, thank you!

Also, aside: are you tied to using an early lint?

I generally recommend writing late lints, unless you're absolutely sure your lint will never need typing and/or name resolution information.

In this case this was just for a small toy example to verify I could get everything running and partially following some documentation on lint passes from the Clippy book since I was just checking for function names in this example. I will keep your advice in mind when writing any non-example lints.

Thanks again for your quick response and taking the time to respond!