Macro use sugg

[DevinR528]

changelog: Add auto applicable suggstion to macro_use_imports

fixes #5275

Where exactly is the wildcard_imports_helper I haven't been able to import anything ex.
use lazy_static; or something like for that I get version/compiler conflicts?
Found it.

Should we also check for #[macro_use] extern crate, although this is still depended on for stuff like rustc_private?

[DevinR528]

This seems to work exactly how we need it to, so far it ignores all the modules the macro they are defined in but not found in. I need to figure out how to import from another crate into my test crate so I can have the macro stick in a module.

Note: since the prelude import is shadowed by the (builtin?) macro prelude import it doesn't catch that any more.

[flip1995]

For macros from another crate you can use the mini-macro subcrate in the Clippy repo: https://github.com/rust-lang/rust-clippy/tree/master/mini-macro

This is also used in some other tests and includes a few proc-macros.

[flip1995]

Since I'm not really sure, how far you're along with implementing this, can you give me a short summary, what you already implemented, what you think needs to be added, and which questions you still might have, before you can continue?

[DevinR528]

It catches anything from any level of an imported crate directly imported

// crate importing from "macros"
extern crate macro_rules;

#[macro_export]
macro_rules! pub_macro {
    () => {
        ()
    };
}
pub mod inner {
    // RE-EXPORT
    // this will stick in `inner` module
    pub use macro_rules::try_err;

    #[macro_export]
    macro_rules! inner_mod {
        () => {
            #[allow(dead_code)]
            pub struct Tardis;
        };
    }
}

// using in this crate 
#[macro_use]
use mac;

fn main() {
    pub_macro!();
    inner_mod!();
    // without the inner:: it fails to compile?
    inner::try_err!();
}

it correctly ignores modules only using the root/crate name, it also aliases correctly so it will use the renamed name. The re-export works (it will produce a suggestion of macro::inner::try_err) but only if I call it like inner::try_err!(), otherwise it fails to compile can't resolve macro try_err. This seems more like me setting up something wrong in the auxiliary crate folder though?

By iterating over the children of an imported Mod we miss std::prelude so the original test of println fails. I'd imagine not many people are #[macro_use] std::prelude so maybe that's ok?

I can update the .stderr file that may make it easier to see what I've done?

[bors]

☔ The latest upstream changes (presumably #5300) made this pull request unmergeable. Please resolve the merge conflicts.

[DevinR528]

Since we know if we encountered a macro that wasn't found in the imports search, and therefor is one of the weird import kinds (std::prelude). Could we defer all the span_lint_and_suggest calls until this point pushing them into a Vec then once we know there are no weird macro imports we emit all the suggestions as machine fixable.

how they are gathered

if lcx.sess().opts.edition == Edition::Edition2018;
if let hir::ItemKind::Use(path, _kind) = &item.kind;
if let Some(mac_attr) = item
    .attrs
    .iter()
    .find(|attr| attr.ident().map(|s| s.to_string()) == Some("macro_use".to_string()));
if let Res::Def(DefKind::Mod, id) = path.res;
then {
    for kid in lcx.tcx.item_children(id).iter() {
        if let Res::Def(DefKind::Macro(_mac_type), mac_id) = kid.res {
            let span = mac_attr.span.clone();
            self.imports.push((lcx.tcx.def_path_str(mac_id), span));
        }
    }
}

where this would happen

fn check_crate_post(&mut self, lcx: &LateContext<'_, '_>, _krate: &hir::Crate<'_>) {
    for (import, span) in self.imports.iter() {
        let matched = self
            .mac_refs
            .iter()
            .find(|(_span, mac)| import.ends_with(&mac.name))
            .is_some();

        if matched {
            self.mac_refs.retain(|(_span, mac)| !import.ends_with(&mac.name));
            let msg = "`macro_use` attributes are no longer needed in the Rust 2018 edition";
            let help = format!("use {}", import);
            span_lint_and_sugg(
                lcx,
                MACRO_USE_IMPORTS,
                *span,
                msg,
                "remove the attribute and import the macro directly, try",
                help,
                Applicability::HasPlaceholders,
            )
        }
    }
    if !self.mac_refs.is_empty() {
        // TODO if not empty we found one we could not make a suggestion for
        // such as std::prelude::v1 or something else I haven't thought of.
        // If we defer the calling of span_lint_and_sugg we can make a decision about its
        // applicability?
    }
}

[bors]

☔ The latest upstream changes (presumably #5319) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #5294) made this pull request unmergeable. Please resolve the merge conflicts.

[flip1995]

@DevinR528 Sorry for taking so long for the review. I don't have an excuse, just wasn't really motivated to review PRs recently. I didn't forget this PR and want to review it in the coming days.

[bors]

☔ The latest upstream changes (presumably #5470) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #5423) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #5141) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #5332) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #5522) made this pull request unmergeable. Please resolve the merge conflicts.

[bors]

☔ The latest upstream changes (presumably #5550) made this pull request unmergeable. Please resolve the merge conflicts.

[DevinR528]

Is this compile failure something I can fix?

error[E0624]: associated function `seek_after` is private
   --> clippy_lints/src/redundant_clone.rs:594:25
    |
594 |         self.maybe_live.seek_after(at);
    |                         ^^^^^^^^^^ private associated function

[flip1995]

#5566 We're currently figuring out some stuff with the move to subtree in the Rust repo, so a rustup may take a day or two more this time. A fix is ready, though.

[flip1995]

stderr differs? That is weird 🤔

[bors]

☔ The latest upstream changes (presumably #5378) made this pull request unmergeable. Please resolve the merge conflicts.

[DevinR528]

It's the ordering of the suggestions, hmmm I'll look into this...

[flip1995]

@bors r+

Thanks!

[bors]

📌 Commit 3f91e39 has been approved by flip1995

[bors]

⌛ Testing commit 3f91e39 with merge 0537c9d...

[bors]

💔 Test failed - checks-action_test

[flip1995]

Wait the problem is, that the output is different on 32bit linux 🤔 Since only the order is different, you can add // ignore-32bit

[flip1995]

You also have to update the .fixed and .stderr files

[DevinR528]

Oops duh 🤦

[flip1995]

Thanks!

@bors r+

[bors]

📌 Commit e521a4e has been approved by flip1995

[bors]

⌛ Testing commit e521a4e with merge f065d4b...

[bors]

☀️ Test successful - checks-action_dev_test, checks-action_remark_test, checks-action_test
Approved by: flip1995
Pushing f065d4b to master...