Add lint to suggest into() to construct reference-counted slice

[rudyardrichter]

To resolve #2623.

[rudyardrichter]

Hi @oli-obk, here's what I started on for #2623. A few questions—how can I print debug information from running the lint? I tried something like this:

cx.sess().span_note(expr.span, &format!("{:?}", ty))

(from that article linked in CONTRIBUTING.md) but it looks like that's outdated. I'm not positive though since it seems like this doesn't actually run the lint at all when I do either of these:

TESTNAME=ui/use_shared_from_slice cargo test --test compile-test

cargo run --bin clippy-driver -- -L ./target/debug tests/ui/use_shared_from_slice.rs

What else do I need to change to get the lint running?

[oli-obk]

The dogfood test complains about

warning: unused import: `LintContext`
 --> clippy_lints/src/use_shared_from_slice.rs:2:57
  |
2 | use rustc::lint::{LateContext, LateLintPass, LintArray, LintContext, LintPass};
  |                                                         ^^^^^^^^^^^
  |
  = note: #[warn(unused_imports)] on by default
error: useless use of `format!`
  --> clippy_lints/src/use_shared_from_slice.rs:59:21
   |
59 |                     format!("TODO"),
   |                     ^^^^^^^^^^^^^^^ help: consider using .to_string(): `"TODO".to_string()`
   |
   = note: `-D useless-format` implied by `-D clippy`
   = help: for further information visit https://rust-lang-nursery.github.io/rust-clippy/v0.0.205/index.html#useless_format

[rudyardrichter]

Sorry, I had committed it after removing some things for debugging. After 5958ddf the tests are passing but I don't think it runs the lint (I don't see anything happen even if I put a panic in check_expr).

[oli-obk]

You need to add the lint to lib.rs.

Also try running the utils/update_lints.py script first

[rudyardrichter]

Yeah I just ran util/update_lints.py and didn't actually change lib.rs myself. It looks correct here

[rudyardrichter]

Ah, sorry, didn't see that running the script didn't add a register_late_lint_pass. Thanks—I'll get back to working on this.

[rudyardrichter]

@oli-obk would you mind checking up on the progress here? I fixed up check_expr and moved the match_type_parameter out from a different lint to the utils to use it for this one, and added the check_local for checking type declarations.

I noticed LateLintPass has a check_struct_field function and thought I might add the same type declaration check for struct fields, but I couldn't find much relevant code or documentation to figure that out, so I've left that out.

[oli-obk]

I have a feeling we're overlooking some cases here. Can you run your lint on a bunch of crates like cargo, rls or other big crates to see what kind of situations trigger your lints there?

[oli-obk]

So this is a little dangerous, because the message might appear even if the user has no way to change the type. For Rc<Vec<[T]>> this is is kinda fine, because the suggestion will still work, but for Rc<String> I'm not sure if .as_str().into() is equivalent.

[phansch]

ping @rudyardrichter, do you still want to finish this up?

[rudyardrichter]

@phansch Yeah, I will get back to this and try running this lint like @oli-obk suggested. I may need some suggestions for how to lint the Rc<String> case

[flip1995]

Ping @rudyardrichter. I'm going over old PRs, that were abandoned by us reviewers (sorry for that!) or by the authors. Are you still interested in completing this?

[rudyardrichter]

Hey @flip1995, probably best to let someone else take this. Apologies.

[flip1995]

No worries. Sorry for not giving you an earlier review. Thanks for your contribution!

[HKalbasi]

finished in #6044
@rustbot label -S-inactive-closed