New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

NLL diagnostics replaced nice closure errors w/ indecipherable free region errors #52572

Merged

bors merged 6 commits into rust-lang:master from davidtwco:issue-51027

Jul 22, 2018

Member

davidtwco commented Jul 20, 2018

Fixes #51027.

r? @nikomatsakis

rust-highfive assigned nikomatsakis

davidtwco changed the title ~~NLL diagnostics replaced nice closure errors w/ indecipherable free region errors~~ WIP: NLL diagnostics replaced nice closure errors w/ indecipherable free region errors

rust-highfive added the S-waiting-on-review label

Member Author

davidtwco commented Jul 20, 2018 •

edited

Loading

As of submission, this PR produces the following error (or equivalent):

error: borrowed data escapes outside of closure
  --> issue-45983.rs:17:27
   |
16 |     let mut x = None;
   |         ----- `x` is declared here, outside of the closure body
17 |     give_any(|y| x = Some(y));:
   |               -  ^^^^^^^^^^^ `y` escapes the closure body here
   |               |
   |               `y` is a reference that is only valid in the closure body

It fixes the following tests mentioned in the issue:

borrowck/issue-45983.rs
borrowck/issue-7573.rs
borrowck/regions-escape-bound-fn-2.rs
borrowck/regions-escape-bound-fn.rs
borrowck/regions-escape-unboxed-closure.rs
closure-expected-type/expect-region-supply-region.rs
error-codes/E0621-does-not-trigger-for-closures.rs (not sure how to approach)

davidtwco mentioned this pull request

NLL diagnostics replaced nice closure errors w/ indecipherable free region errors #51027

Closed

7 tasks

davidtwco commented

View reviewed changes

src/librustc/mir/tcx.rs Outdated

    
                                      }

                                  },

                                  ProjectionElem::Deref => {

                                      if recurse {

Member Author

davidtwco Jul 20, 2018 •

edited

Loading

I wasn't sure about adding this parameter (recurse) - without it, a handful of unrelated tests start throwing duplicate errors with shorter spans, but without considering the place within the deref I don't get some of the tests that currently work.

Contributor

nikomatsakis Jul 21, 2018

Hmm. So the behavior when recurse is true seems mostly correct to me -- that is, if you have a variable that is captured "by reference" (and not "by value"), then x would be desugared to *self.x. However, I think we would only then want to recurse through a single deref. Maybe this is the cause of the other tests behaving strangely?

Member Author

davidtwco Jul 21, 2018

I've just checked and it seems like only doing a single deref still exhibits this behaviour.

Member Author

davidtwco Jul 21, 2018

Pushed a fix for this.

nikomatsakis requested changes

View reviewed changes

src/librustc_mir/borrow_check/nll/region_infer/error_reporting/var_name.rs Outdated

    
                              self.get_argument_index_for_region(tcx, fr)

                                  .map(|index| self.get_argument_name_and_span_for_region(mir, index))

                          })

                          .unwrap_or_else(|| span_bug!(mir.span, "can't find var name for free region {:?}", fr))

Contributor

nikomatsakis Jul 21, 2018

I'm a bit nervous about this case. Based on the other code I see, it seems like it is possible for the region to appear in the closure output, for example. I'll try to build your branch and produce an example though because it's hard for me to predict in advance.

Member Author

davidtwco Jul 21, 2018

Most recent pushed commit introduces changes that ICE due to this in some tests.

Member Author

davidtwco Jul 21, 2018

Another commit now stops those ICEs by falling back to previous error handling.

src/librustc_mir/borrow_check/nll/region_infer/error_reporting/var_name.rs Outdated

    
                      (Some(upvar_name), upvar_span)

                  }

                  /// Get argument index for a region.

Contributor

nikomatsakis Jul 21, 2018

I would word this as:

Search the argument types for one that references fr (which should be a free region). Returns Some(_) with the index of the input if one is found.

NB: In the case of a clsoure, the index is indexing into the signature as seen by the user--- in particular, index 0 is not the implicit self parameter.

Member Author

davidtwco Jul 21, 2018

Pushed a commit that addresses this.

src/librustc_mir/borrow_check/nll/region_infer/error_reporting/var_name.rs Outdated

    
                      Some(upvar_index)

                  }

                  /// Get upvar name and span for a region.

Contributor

nikomatsakis Jul 21, 2018

I would say something like:

Given the index of an upvar, finds its name (if any) and the span from where it was declared.

Question: Why does this return an Option? From what I can tell, it is always Some?

Member Author

davidtwco Jul 21, 2018 •

edited

Loading

I had it return an Option so that the types where the same in the and_then sections of the function above. I'll change this to use a .map(...) in that function rather than change the return type of this one.

Member Author

davidtwco Jul 21, 2018

Pushed a commit that addresses this.

src/librustc_mir/borrow_check/nll/region_infer/error_reporting/var_name.rs Outdated

    
                          .unwrap_or_else(|| span_bug!(mir.span, "can't find var name for free region {:?}", fr))

                  }

                  /// Get upvar index for a region.

Contributor

nikomatsakis Jul 21, 2018

I would say something like:

Search the upvars (if any) to find one that references fr. Return its index.

Member Author

davidtwco Jul 21, 2018

Pushed a commit that addresses this.

src/librustc_mir/borrow_check/nll/region_infer/error_reporting/var_name.rs Outdated

    
                      Some(argument_index)

                  }

                  /// Get argument name and span for a region.

Contributor

nikomatsakis Jul 21, 2018

I would reword this similar to the fn for upvars.

Member Author

davidtwco Jul 21, 2018

Pushed a commit that addresses this.

src/librustc_mir/borrow_check/nll/region_infer/error_reporting/mod.rs Outdated

    
            @@ -224,30 +235,75 @@ impl<'tcx> RegionInferenceContext<'tcx> {
          
                      // Get a span

                      let (category, span) = categorized_path.first().unwrap();

                      match category {

                          ConstraintCategory::AssignmentToUpvar =>

Contributor

nikomatsakis Jul 21, 2018

This is interesting. Looking for an assignment to an upvar is not quite taking the approach I expected, but it also makes sense. I guess though that this code won't fire for something like:

fn foo(x: &mut Vec<&u32>) {
  bar(|y| x.push(y));
}

fn bar(_: impl FnOnce(&u32)) { }

Right?

Contributor

nikomatsakis Jul 21, 2018

Seems ok, though I can think we can address this case via a similar set of errors, if we generalize the conditions somewhat away from assignment and more into looking at the kinds of regions involved. (Or maybe this is just another sort of case to specialize)

Member Author

davidtwco Jul 21, 2018

This is similar to the issue-7573.rs case that isn't yet handled. I messed around with making another variant for ConstraintCategory that would go through to the report_closure_error branch.

Member Author

davidtwco Jul 21, 2018

Fixed with most recent commit.

nikomatsakis changed the title ~~WIP: NLL diagnostics replaced nice closure errors w/ indecipherable free region errors~~ NLL diagnostics replaced nice closure errors w/ indecipherable free region errors

nikomatsakis approved these changes

View reviewed changes

Contributor

nikomatsakis commented Jul 22, 2018

Contributor

bors commented Jul 22, 2018

📌 Commit 58eee34fcc9e3642942690b979d78a7002ddaf7b has been approved by nikomatsakis

bors added S-waiting-on-bors and removed S-waiting-on-review labels

Contributor

bors commented Jul 22, 2018

☔ The latest upstream changes (presumably #52359) made this pull request unmergeable. Please resolve the merge conflicts.

bors added S-waiting-on-author and removed S-waiting-on-bors labels

davidtwco added 6 commits

July 22, 2018 12:23


          Classify aggregate rvalues as assignments.

24c5751


          Improved closure errors.

571eec6


          Improved documentation of functions in new module.

f71dbbb


          Improved is_upvar_field_projection - no longer need recurse parameter.

ce4f446


          Modified how constraint classification happens to upvars, can now han…

b6dfa8c

…dle function call case.


          Fallback to general error handling in ICE cases.

c64db00

davidtwco force-pushed the issue-51027 branch from 58eee34 to c64db00 Compare

July 22, 2018 12:43

Contributor

nikomatsakis commented Jul 22, 2018

Contributor

bors commented Jul 22, 2018

📌 Commit c64db00 has been approved by nikomatsakis

bors added S-waiting-on-bors and removed S-waiting-on-author labels

Contributor

nikomatsakis commented Jul 22, 2018

Contributor

nikomatsakis commented Jul 22, 2018

Giving high priority, as improving this case is a EP2 blocker.

Contributor

bors commented Jul 22, 2018

⌛ Testing commit c64db00 with merge 32772fd...

bors added a commit that referenced this pull request


          Auto merge of #52572 - davidtwco:issue-51027, r=nikomatsakis

32772fd

NLL diagnostics replaced nice closure errors w/ indecipherable free region errors

Fixes #51027.

r? @nikomatsakis

Contributor

bors commented Jul 22, 2018

☀️ Test successful - status-appveyor, status-travis
Approved by: nikomatsakis
Pushing 32772fd to master...

bors merged commit c64db00 into rust-lang:master

davidtwco deleted the issue-51027 branch

July 22, 2018 19:30

pnkfelix mentioned this pull request

error: internal compiler error: Accessing (*_310) with the kind Write(Move) shouldn't be possible #54597

Closed

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

S-waiting-on-bors