Rework "long type names" printing logic #136328

estebank · 2025-01-31T03:27:35Z

Make it so more type-system types can be printed in a shortened version (like Predicates).

Centralize printing the information about the "full type name path".

Make the "long type path" for the file where long types are written part of Diag, so that it becomes easier to keep track of it, and ensure it will always will be printed out last in the diagnostic by making its addition to the output implicit.

Tweak the shortening of types in "expected/found" labels.

Remove dead file note.rs.

Make it so more type-system types can be printed in a shortened version (like `Predicate`s). Centralize printing the information about the "full type name path". Make the "long type path" for the file where long types are written part of `Diag`, so that it becomes easier to keep track of it, and ensure it will always will be printed out last in the diagnostic by making its addition to the output implicit. Tweak the shortening of types in "expected/found" labels. Remove dead file `note.rs`.

rustbot · 2025-01-31T03:27:43Z

r? @spastorino

rustbot has assigned @spastorino.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

rustbot · 2025-01-31T03:27:45Z

Some changes occurred in match checking

cc @Nadrieril

Some changes occurred in need_type_info.rs

cc @lcnr

estebank · 2025-01-31T03:28:46Z

compiler/rustc_middle/src/ty/error.rs

+    pub fn short_string<'a, T>(self, p: T, path: &mut Option<PathBuf>) -> String
+    where
+        T: Print<'tcx, FmtPrinter<'a, 'tcx>> + Lift<TyCtxt<'tcx>> + Copy + Hash,
+        <T as Lift<TyCtxt<'tcx>>>::Lifted: Print<'tcx, FmtPrinter<'a, 'tcx>>,
+    {


With these changes now anything that can be printed can be passed to this, and the printing logic already handles the shortening gracefully. Look at the long E0277 error for an example of the effect.

estebank · 2025-01-31T03:29:51Z

compiler/rustc_errors/src/diagnostic.rs

    fn take_diag(&mut self) -> DiagInner {
+        if let Some(path) = &self.long_ty_path {
+            self.note(format!(
+                "the full name for the type has been written to '{}'",
+                path.display()
+            ));
+            self.note("consider using `--verbose` to print the full type name to the console");
+        }
        Box::into_inner(self.diag.take().unwrap())
    }


Now there's no chance of forgetting to print this path (although we need to remember to use long_ty_path() to set the path right after error construction if we've made short strings before that).

estebank · 2025-01-31T03:30:45Z

compiler/rustc_borrowck/src/diagnostics/move_errors.rs

@@ -635,19 +629,13 @@ impl<'infcx, 'tcx> MirBorrowckCtxt<'_, 'infcx, 'tcx> {
                    self.suggest_cloning(err, place_ty, expr, Some(use_spans));
                }

-                let mut path = None;
-                let ty = self.infcx.tcx.short_ty_string(place_ty, &mut path);
+                let ty = self.infcx.tcx.short_string(place_ty, err.long_ty_path());


This is an example of the intended way of using this. This is the safest way of creating a short string because there's no way of forgetting to print the "full name written to path" message.

estebank · 2025-01-31T03:31:53Z

compiler/rustc_middle/src/ty/print/pretty.rs

+                    if self.should_truncate() {
+                        write!(self, "@...}}")?;
+                        return Ok(());


This changes the closure printing so when the closure is right at the edge of being replaced with ..., we print it as {closure@...}.

estebank · 2025-01-31T03:32:51Z

compiler/rustc_trait_selection/src/error_reporting/infer/note.rs

This file was in the repo but wasn't modded 😵

estebank · 2025-01-31T03:33:36Z

tests/ui/coercion/coerce-expect-unsized-ascribed.stderr

@@ -29,7 +29,7 @@ error[E0308]: mismatched types
  --> $DIR/coerce-expect-unsized-ascribed.rs:14:27
   |
 LL |     let _ = type_ascribe!(Box::new( { |x| (x as u8) }), Box<dyn Fn(i32) -> _>);
-   |                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `Box<dyn Fn(i32) -> u8>`, found `Box<{closure@coerce-expect-unsized-ascribed.rs:14:39}>`
+   |                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `Box<dyn Fn(i32) -> u8>`, found `Box<{closure@...}>`


Example of the "short closure" logic mentioned before.

estebank · 2025-01-31T03:34:10Z

tests/ui/diagnostic-width/E0271.ascii.stderr

@@ -7,14 +7,16 @@ LL | |             Err::<(), _>(
 LL | |                 Ok::<_, ()>(
 ...  |
 LL | |     )
-   | |_____^ type mismatch resolving `<Result<Result<(), Result<Result<(), ...>, ...>>, ...> as Future>::Error == Foo`
+   | |_____^ type mismatch resolving `<Result<..., ()> as Future>::Error == Foo`


Example of the allowlisting of "known short" types so we don't replace them with ... unnecessarily.

estebank · 2025-01-31T03:35:22Z

tests/ui/diagnostic-width/long-e0277.stderr

+error[E0277]: the trait bound `(..., ..., ..., ...): Trait` is not satisfied
+  --> $DIR/long-e0277.rs:13:21
+   |
+LL |     require_trait::<D>();
+   |                     ^ unsatisfied trait bound
+   |
+   = help: the trait `Trait` is not implemented for `(..., ..., ..., ...)`
+help: this trait has no implementations, consider adding one
+  --> $DIR/long-e0277.rs:8:1
+   |
+LL | trait Trait {}
+   | ^^^^^^^^^^^
+note: required by a bound in `require_trait`
+  --> $DIR/long-e0277.rs:10:21
+   |
+LL | fn require_trait<T: Trait>() {}
+   |                     ^^^^^ required by this bound in `require_trait`
+   = note: the full name for the type has been written to '$TEST_BUILD_DIR/diagnostic-width/long-e0277/long-e0277.long-type-hash.txt'
+   = note: consider using `--verbose` to print the full type name to the console


This is the error I had set out to clean up, and ended up being a slightly larger refactor than I intended. You can see that the bound is reusing the "short type" printing logic now.

estebank · 2025-01-31T03:36:22Z

tests/ui/traits/on_unimplemented_long_types.stderr

-   = note: the full name for the type has been written to '$TEST_BUILD_DIR/traits/on_unimplemented_long_types/on_unimplemented_long_types.long-type-hash.txt'
-   = note: consider using `--verbose` to print the full type name to the console


Another benefit of making Diag carry the path: it's impossible to have redundant notes now.

estebank · 2025-01-31T03:45:16Z

#136315 will conflict with this PR.

lqd · 2025-01-31T06:54:03Z

tests/ui/diagnostic-width/E0271.rs

@@ -1,6 +1,6 @@
 //@ revisions: ascii unicode
-//@[ascii] compile-flags: --diagnostic-width=40
-//@[unicode] compile-flags: -Zunstable-options --error-format=human-unicode --diagnostic-width=40
+//@[ascii] compile-flags: --diagnostic-width=40 -Zwrite-long-types-to-disk=yes


Random comment while I was in the neighborhood.

When using -Zwrite-long-types-to-disk=yes, we usually normalize away the hash in the long type file name, but that's not enough. Using a compare-mode will embed the compare mode in the long type file path as well.

The path note: the full name for the type has been written to '$TEST_BUILD_DIR/diagnostic-width/E0271.unicode/E0271.long-type-hash.txt' will become note: the full name for the type has been written to '$TEST_BUILD_DIR/diagnostic-width.$compare-mode/E0271.unicode/E0271.long-type-hash.txt' and the test will fail.

I encountered this while looking over the test failures under the new solver, i.e. ./x test bla --compare-mode=next-solver. While this specific test will currently fail for different reasons, it should pass in general without any changes required to diagnostics. (you can check that this test will fail under e.g. --compare-mode polonius)

For that, we'd need either:

compiletest to take care of all normalizations needing to be done due to compare-modes

expand what we manually normalize in the tests, taking care of the compare-mode suffix, as well as the possible hash

Since we already do part of the latter, I personally went with option 2 in #136310, and expanded the regex to look for the complete file path to remove the subdirectories.

For cases that have a hash suffix to normalize away, like in this E0271 test, it looks like:

// The regex below normalizes the long type file name to make it suitable for compare-modes. //@ normalize-stderr: "'\$TEST_BUILD_DIR/.*\.long-type-\d+.txt'" -> "'$$TEST_BUILD_DIR/$$FILE.long-type-hash.txt'"

One could remove the comment to avoid reblessing the new line numbers.

I already took care of long-E0308 and so on in that other PR (so they will conflict/require reblessing), but the same issue will also appear in the new long-e0277.rs test here.

Thanks for coming to my TED talk.

rustbot assigned spastorino Jan 31, 2025

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jan 31, 2025

estebank commented Jan 31, 2025

View reviewed changes

This was referenced Jan 31, 2025

Long type name emitted if the type name is part of a suggestion #135919

Open

Use short ty string for binop errors #136315

Open

jieyouxu self-assigned this Jan 31, 2025

lqd reviewed Jan 31, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Rework "long type names" printing logic #136328

Rework "long type names" printing logic #136328

estebank commented Jan 31, 2025

rustbot commented Jan 31, 2025

rustbot commented Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank Jan 31, 2025

estebank commented Jan 31, 2025

lqd Jan 31, 2025

		= note: the full name for the type has been written to '$TEST_BUILD_DIR/traits/on_unimplemented_long_types/on_unimplemented_long_types.long-type-hash.txt'
		= note: consider using `--verbose` to print the full type name to the console

Rework "long type names" printing logic #136328

Are you sure you want to change the base?

Rework "long type names" printing logic #136328

Conversation

estebank commented Jan 31, 2025

rustbot commented Jan 31, 2025

rustbot commented Jan 31, 2025

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

estebank commented Jan 31, 2025

Choose a reason for hiding this comment