Allow simultaneous usage of `EarlyLintPass` and `LateLintPass`

[CleverWild]

I've seen this pattern in some clippy lints, and I don't have any idea what the reason for this "restriction" is.

Why do I ask?
I am writing now a lint that checks the path value from dropshot::endpoint which is proc-macro-attribute and warns when a user tries to use bad patterns, e.g., passes the ID value in json instead of the path. In EarlyLintPass I have access to ast, where the attribute is located. But I don't have access to hir type info to check argument types for some pattern similar to passing the ID value in json. Whereas LateLintPass has the opposite situation.

[smoelius]

I don't think there is an easy way to solve your problem as written. The EarlyLintPass and LateLintPass lints are run in separate passes in the compiler. So, allowing their simultaneous usage would require compiler changes.

FWIW, I have seen many people express frustration over this. It is common to operate on the HIR, to want something in the AST, and to not have easy access to it.

But regarding your specific problem, this what you are processing? https://doc.rust-lang.org/nightly/nightly-rustc/rustc_hir/attrs/enum.AttributeKind.html#variant.ProcMacroAttribute

Could you extract the syntax using snippet_opt and then parse it using syn?

That would be a less-than-ideal solution because you are bringing in another parser (syn). But would that work?

[CleverWild]

I don't think there is an easy way to solve your problem as written. The EarlyLintPass and LateLintPass lints are run in separate passes in the compiler. So, allowing their simultaneous usage would require compiler changes.

See https://github.com/rust-lang/rust-clippy/blob/master/clippy_lints/src/large_include_file.rs

The solution I see is to write an additional macro, similar to declare_early_lint, but which registers both early and late checks.

[smoelius]

The solution I see is to write an additional macro, similar to declare_early_lint, but which registers both early and late checks.

OK. I think that is probably doable. I'm going to have to give it some thought, though. I think this is the code that will have to change: https://docs.rs/dylint_linting/4.1.2/src/dylint_linting/lib.rs.html#271-293

I'd be open to reviewing a PR if someone wanted to tackle it.