Add functionality for epoch lints; add epoch lint for dyn-trait #48461

Manishearth · 2018-02-23T08:25:41Z

This makes it possible for lints to specify an epoch after which they become Deny, as well as automatically creating lint groups for these.

This had to be done as a hardwired lint because we lose information about whether TyTraitObject nodes were from dyn trait in the HIR, and in the AST we don't know if TyPaths are type-y or trait-y.

Still need to write tests and fix up all the new warnings (though I may wait for contributors to do that -- @killercup is it possible to run rustfix on a crate-by-crate basis for rustc?). Also, #48457 needs to get some contents.

Still, worth getting review for what I have so far.

r? @nmatsakis

fixes #48447

Manishearth · 2018-02-23T08:28:56Z

src/librustc_lint/lib.rs

+         FutureIncompatibleInfo {
+             id: LintId::of(lint::builtin::BARE_TRAIT_OBJECT),
+             reference: "issue #48457 <https://github.com/rust-lang/rust/issues/48457>",
+             epoch: Some(session::config::Epoch::Epoch2018),


we could potentially reuse the epoch on the lint

we do need something for lints which need to be removed on the next epoch (because the stuff they lint doesn't exist anymore)

I'm trying to imagine when we would want to make a distinction between the "epoch on the lint" and the "epoch in the future compat info". I'm not really sure -- indeed -- what the "epoch on the lint" means, if not that it is an epoch lint of this kind. Maybe another way of saying this is that "future incompatibility" feels distinct from epochs, and might want to be a different mechanism. In particular, future incompatible changes change for everyone, whereas epochs are opt-in.

The lint may become Deny on a different epoch than the one it is linting for, basically.

One of the settings is "when this becomes Deny", the other is "which lint group this belongs to".

In the future we may also provide a way for the lint to go away completely.

Manishearth · 2018-02-23T08:35:22Z

We also need a lint for the dyn keyword, yes? I can't write this one until we have raw identifiers.

oli-obk · 2018-02-23T09:17:38Z

src/librustc/lint/levels.rs

@@ -1,4 +1,4 @@
-// Copyright 2017 The Rust Project Developers. See the COPYRIGHT
+// Copyright 2017 The Rust Project ~velopers. See the COPYRIGHT


accidental replace

oli-obk · 2018-02-23T09:19:36Z

src/librustc/session/mod.rs

+    }
+
+    pub fn buffer_lint_with_diagnostic<S: Into<MultiSpan>>(&self,
+                                           lint: &'static lint::Lint,


indentation is weird here. Maybe use rustfmt style?

oli-obk · 2018-02-23T09:23:15Z

src/librustc/lint/mod.rs

@@ -74,25 +75,46 @@ pub struct Lint {
    ///
    /// e.g. "imports that are never used"
    pub desc: &'static str,
+
+    /// Deny lint after this epoch
+    pub epoch_deny: Option<config::Epoch>,


Further down you state that the lint should be removed after the epoch, here it seems to suggest it should be deny.

Where is this? It used to say this, but that changed.

This specific lint becomes Deny in Rust2018. Other lints may wish to disappear entirely because they are no longer lints.

Manishearth · 2018-02-23T18:04:04Z

A bug is that it will currently suggest &dyn ::Foo for &::Foo, however dyn::Foo parses as a path. We could suggest &dyn (::Foo) (at least on the 2015 epoch) but this doesn't currently work either.

Manishearth · 2018-02-23T18:31:26Z

I've got the rustfix stuff running at #48477

Manishearth · 2018-02-23T20:57:09Z

#48481 fixes the dyn (Trait) issue

nikomatsakis

This all looks quite good. I'm wondering about the idea of distinguish epochal vs non-epochal future incompatibility. Thoughts?

r=me modulo nits and this question though.

nikomatsakis · 2018-02-27T19:25:01Z

src/librustc/lint/context.rs

-        for info in lints {
-            self.future_incompatible.insert(info.id, info);
+
+        for epoch in config::ALL_EPOCHS {


comment plz

"Create a lint group for each epoch corresponding to the future compatibility warnings targeting that epoch"

(Do you think that this ought to include the lints for "all previous epochs", since they are additive?)

(Also, are these lint groups insta-stable?)

We could. I think having them separate is cleaner, Rust doesn't do well with overlapping lint groups.

Rust lint stability is a very low bar, you just need to not remove the lint (aside from register_removed_lint() or whatever it's called). I think it's fine to insta-stable lint groups.

nikomatsakis · 2018-02-27T19:26:53Z

src/librustc_lint/lib.rs

+         FutureIncompatibleInfo {
+             id: LintId::of(lint::builtin::BARE_TRAIT_OBJECT),
+             reference: "issue #48457 <https://github.com/rust-lang/rust/issues/48457>",
+             epoch: Some(session::config::Epoch::Epoch2018),


I'm trying to imagine when we would want to make a distinction between the "epoch on the lint" and the "epoch in the future compat info". I'm not really sure -- indeed -- what the "epoch on the lint" means, if not that it is an epoch lint of this kind. Maybe another way of saying this is that "future incompatibility" feels distinct from epochs, and might want to be a different mechanism. In particular, future incompatible changes change for everyone, whereas epochs are opt-in.

nikomatsakis · 2018-02-27T19:30:33Z

src/librustc/lint/builtin.rs

-                    Err(_) => format!("dyn <type>")
+            BuiltinLintDiagnostics::BareTraitObject(span, is_global) => {
+                let sugg = match (sess.codemap().span_to_snippet(span), is_global) {
+                    (Ok(s), true) => format!("dyn ({})", s),


Nit: this would (imo) be slightly easier to read as

let sugg = match sess.codemap().span_to_snippet(span) { Ok(s) if is_global => ... Ok(s) => ... _ }

We do that, perhaps you're reviewing an earlier commit?

nikomatsakis · 2018-02-27T19:31:36Z

src/librustc/lint/builtin.rs

+                    Ok(s) => format!("dyn {}", s),
+                    Err(_) => format!("dyn <type>")
+                };
+                db.span_suggestion(span, "use `dyn`", sugg);


can we have a test or two, particularly of the suggestions?

Manishearth · 2018-02-28T03:24:45Z

@bors r=nmatsakis

(r+d over IRC)

bors · 2018-02-28T03:24:46Z

📌 Commit 0cb3672 has been approved by nmatsakis

@nrc

…sakis Fixes rust-lang#47311. r? @nrc

Rollup of 10 pull requests - Successful merges: #48355, #48359, #48380, #48419, #48420, #48461, #48522, #48570, #48572, #48603 - Failed merges:

steveklabnik · 2018-03-01T15:26:48Z

(r+d over IRC)

@Manishearth any chance you could link the point in the log? It sounds like you two discussed some things, the justification would be nice to have saved in-context in the PR. :)

Manishearth · 2018-03-01T16:01:09Z

https://botbot.me/mozilla/rust-internals/2018-02-28/?msg=97307371&page=1

…sakis Use `dyn trait` everywhere Based on rust-lang#48461 do not land unless both are reviewed.

Add ToString and FromStr impls for Epoch

d9438c3

Manishearth commented Feb 23, 2018

View reviewed changes

Manishearth force-pushed the epoch-dyn-trait branch from c618b0f to 76a2705 Compare February 23, 2018 08:29

oli-obk reviewed Feb 23, 2018

View reviewed changes

Manishearth added 5 commits February 23, 2018 08:24

Allow future-incompat lints to mention an epoch

da9dc05

Add hardwired lint for dyn trait

3eeabe7

Add ability for hardwired lints to operate on the diagnostic builder

bd29696

Lint bare traits

63168f7

span_bug doesn't work well at this stage, use the session directly

177271f

Manishearth force-pushed the epoch-dyn-trait branch from 76a2705 to 177271f Compare February 23, 2018 16:36

Manishearth mentioned this pull request Feb 23, 2018

Use dyn trait everywhere #48477

Merged

Silence warning in test

dd67fe1

Manishearth force-pushed the epoch-dyn-trait branch 4 times, most recently from 5484cf9 to 1c99be5 Compare February 23, 2018 21:56

nikomatsakis self-assigned this Feb 23, 2018

pietroalbini added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Feb 26, 2018

nikomatsakis approved these changes Feb 27, 2018

View reviewed changes

Manishearth force-pushed the epoch-dyn-trait branch 6 times, most recently from 6ac8093 to a4ebd4a Compare February 28, 2018 03:08

Emit parentheses in suggestion for global paths

0cb3672

Manishearth force-pushed the epoch-dyn-trait branch from a4ebd4a to 0cb3672 Compare February 28, 2018 03:24

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Feb 28, 2018

Manishearth mentioned this pull request Feb 28, 2018

Rollup of 10 pull requests #48615

Merged

Manishearth added a commit to Manishearth/rust that referenced this pull request Feb 28, 2018

Rollup merge of rust-lang#48461 - Manishearth:epoch-dyn-trait, r=nmat…

f57835b

…sakis Fixes rust-lang#47311. r? @nrc

bors added a commit that referenced this pull request Feb 28, 2018

Auto merge of #48615 - Manishearth:rollup, r=Manishearth

0e3c9bb

Rollup of 10 pull requests - Successful merges: #48355, #48359, #48380, #48419, #48420, #48461, #48522, #48570, #48572, #48603 - Failed merges:

bors merged commit 0cb3672 into rust-lang:master Mar 1, 2018

petrochenkov mentioned this pull request Mar 1, 2018

Turn feature-gate table into a query so it is covered by dependency tracking. #48208

Merged

Manishearth deleted the epoch-dyn-trait branch March 1, 2018 16:01

Manishearth added a commit to Manishearth/rust that referenced this pull request Mar 3, 2018

Rollup merge of rust-lang#48477 - Manishearth:dyn-trait-fixes, r=nmat…

d54cf1b

…sakis Use `dyn trait` everywhere Based on rust-lang#48461 do not land unless both are reviewed.

killercup mentioned this pull request Mar 21, 2018

Provide rustfix to upgrade code to Rust 2018 automatically #49247

Closed

8 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add functionality for epoch lints; add epoch lint for dyn-trait #48461

Add functionality for epoch lints; add epoch lint for dyn-trait #48461

Manishearth commented Feb 23, 2018 •

edited

Loading

Manishearth Feb 23, 2018

nikomatsakis Feb 27, 2018

Manishearth Feb 27, 2018

Manishearth commented Feb 23, 2018

oli-obk Feb 23, 2018

oli-obk Feb 23, 2018

oli-obk Feb 23, 2018

Manishearth Feb 23, 2018

Manishearth commented Feb 23, 2018 •

edited

Loading

Manishearth commented Feb 23, 2018 •

edited

Loading

Manishearth commented Feb 23, 2018

nikomatsakis left a comment

nikomatsakis Feb 27, 2018

Manishearth Feb 27, 2018

nikomatsakis Feb 27, 2018

nikomatsakis Feb 27, 2018

Manishearth Feb 27, 2018

nikomatsakis Feb 27, 2018

Manishearth commented Feb 28, 2018

bors commented Feb 28, 2018

steveklabnik commented Mar 1, 2018

Manishearth commented Mar 1, 2018

		@@ -1,4 +1,4 @@
		// Copyright 2017 The Rust Project Developers. See the COPYRIGHT
		// Copyright 2017 The Rust Project ~velopers. See the COPYRIGHT

Add functionality for epoch lints; add epoch lint for dyn-trait #48461

Add functionality for epoch lints; add epoch lint for dyn-trait #48461

Conversation

Manishearth commented Feb 23, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Manishearth commented Feb 23, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Manishearth commented Feb 23, 2018 • edited Loading

Manishearth commented Feb 23, 2018 • edited Loading

Manishearth commented Feb 23, 2018

nikomatsakis left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Manishearth commented Feb 28, 2018

bors commented Feb 28, 2018

steveklabnik commented Mar 1, 2018

Manishearth commented Mar 1, 2018

Manishearth commented Feb 23, 2018 •

edited

Loading

Manishearth commented Feb 23, 2018 •

edited

Loading

Manishearth commented Feb 23, 2018 •

edited

Loading