normalize_doc_attributes + wrap_comments can leave trailing whitespace

Summary

I tried to format this code:

// rustfmt-normalize_doc_attributes: true
// rustfmt-wrap_comments: true

pub trait ClientSystemAlertsExt {
    #[doc = "If this is included, only alerts created at or after this time will be returned.\n```ignore\nx\n```"]
    fn alert_list(&self) -> builder::AlertList<'_>;
}

Expected behavior

No trailing whitespace in any line (and no changes after running rustfmt on the formatted code i.e. idempotency).

Actual behavior

Trailing whitespace after " returned."

Configuration

As noted above: normalize_doc_attributes + wrap_comments

Seems related to #6695

Confirmed I can reproduce this with: rustfmt 1.10.0-nightly (138684c 2026-08-25)

cargo run --bin rustfmt -- --config=normalize_doc_attributes=true,wrap_comments=true

Input

pub trait ClientSystemAlertsExt {
    #[doc = "If this is included, only alerts created at or after this time will be returned.\n```ignore\nx\n```"]
    fn alert_list(&self) -> builder::AlertList<'_>;
}

Output

pub trait ClientSystemAlertsExt {
    ///If this is included, only alerts created at or after this time will be
    /// returned. 
    /// ```ignore
    /// x
    /// ```
    fn alert_list(&self) -> builder::AlertList<'_>;
}

error[internal]: left behind trailing whitespace
 --> <stdin>:6:6:18
  |
6 |     /// returned. 
  |                  ^
  |
  = note: set `error_on_unformatted = false` to suppress the warning against comments or string literals

warning: rustfmt has failed to format. See previous 1 errors.

I have a fix that doesn't regress anything else in the test suite:

--- a/src/comment.rs
+++ b/src/comment.rs
@@ -894,6 +894,11 @@ impl<'a> CommentRewrite<'a> {
                 self.result.pop();
             }
             if self.code_block_attr.is_some() && self.is_prev_line_multi_line {
+                // The fence opens on its own line, so drop the space that was added to
+                // continue the previous wrapped line.
+                if self.result.ends_with(' ') {
+                    self.result.pop();
+                }
                 self.result.push_str(&self.comment_line_separator);
             }
             self.result.push_str(line);

Happy to submit a PR if that would be useful.

@ahl

@ahl instead of popping off a single whitespace character I think we should understand why it was incorrectly added in the first place. We should prevent that from happening instead.

I tried that first because it's a pattern I see quite a bit. How about this instead?

--- a/src/comment.rs
+++ b/src/comment.rs
@@ -818,7 +818,10 @@ impl<'a> CommentRewrite<'a> {
             if line.is_empty() {
                 return false;
             }
-        } else if self.is_prev_line_multi_line && !line.is_empty() {
+        } else if self.is_prev_line_multi_line
+            && !line.is_empty()
+            && !self.code_block_attr.is_some()
+        {
             self.result.push(' ')
         } else if is_last && line.is_empty() {
             // trailing blank lines are unwanted
@@ -893,9 +896,6 @@ impl<'a> CommentRewrite<'a> {
                 // Remove space if this is an empty comment or a doc comment.
                 self.result.pop();
             }
-            if self.code_block_attr.is_some() && self.is_prev_line_multi_line {
-                self.result.push_str(&self.comment_line_separator);
-            }
             self.result.push_str(line);
             self.fmt.shape = Shape::legacy(self.max_width, self.fmt_indent);
             self.is_prev_line_multi_line = false;

I don't fully understand the rationale for the block of code I deleted, but it seems to only impact this one case now and deleting it is an improvement.

@ahl

I don't fully understand the rationale for the block of code I deleted, but it seems to only impact this one case now and deleting it is an improvement.

@ahl check the git history to figure out why that was added?

Looks liked a related case: #6417

@ytmimi

@ytmimi please let me know if a PR along these lines would be reasonable.

Yeah sure, let's go for it. Let's include the current example as a test case, but let's also see if we need to add any others to catch edge cases based on these code changes.

Navigation Menu

Repository navigation

normalize_doc_attributes + wrap_comments can leave trailing whitespace #7069

Summary

Expected behavior

Actual behavior

Configuration

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

normalize_doc_attributes + wrap_comments can leave trailing whitespace #7069

Description

Summary

Expected behavior

Actual behavior

Configuration

Activity

ytmimi commented on Aug 25, 2026

ytmimi commented on Aug 25, 2026

ahl commented on Aug 25, 2026

ytmimi commented on Aug 25, 2026

ahl commented on Aug 25, 2026

ytmimi commented on Aug 25, 2026

ahl commented on Aug 25, 2026

ahl commented on Sep 2, 2026

ytmimi commented on Sep 2, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions