libs/resolver/display.rs (1)

67-111: Multiple hunks are merged into one block and line numbers become incorrect

handle_diff appends all hunks’ removed/added lines into self.orig / self.edit and only calls flush_changes() once at the end. That has two concrete effects:

• Separate change regions (distinct hunks from diff.hunks()) are rendered as a single diff block instead of separate regions.
• orig_line / edit_line never get updated per hunk, so all lines in the combined block are numbered as if they started at line 1, which is wrong for files with multiple distant changes.

Given imara-diff’s Hunk.before / Hunk.after are 0-based index ranges into input.before / input.after (one token per line), you can restore correct behavior by flushing per hunk and resetting the starting line numbers from those ranges. For example:

-  fn handle_diff(mut self, diff: Diff) -> String {
-    for hunk in diff.hunks() {
-      if !hunk.before.is_empty() || !hunk.after.is_empty() {
-        self.has_changes = true;
-      }
-      // collect all removed lines
-      for (i, del) in hunk.before.enumerate() {
-        let s = self.input.interner[self.input.before[del as usize]];
-        if i > 0 {
-          self.orig.push('\n');
-        }
-        self.orig.push_str(&fmt_rem_text_highlight(s));
-      }
-      // collect all added lines
-      for (i, ins) in hunk.after.enumerate() {
-        let s = self.input.interner[self.input.after[ins as usize]];
-        if i > 0 {
-          self.edit.push('\n');
-        }
-        self.edit.push_str(&fmt_add_text_highlight(s));
-      }
-    }
-
-    self.flush_changes();
-    self.output
-  }
+  fn handle_diff(mut self, diff: Diff) -> String {
+    let mut first = true;
+
+    for hunk in diff.hunks() {
+      if hunk.before.is_empty() && hunk.after.is_empty() {
+        continue;
+      }
+
+      // flush previous hunk’s block before starting a new one
+      if !first {
+        self.flush_changes();
+      } else {
+        first = false;
+      }
+
+      // reset starting line numbers from absolute token indices
+      self.orig_line = hunk.before.start as usize + 1;
+      self.edit_line = hunk.after.start as usize + 1;
+      self.has_changes = true;
+
+      // collect removed lines for this hunk
+      for (i, del) in hunk.before.enumerate() {
+        let s = self.input.interner[self.input.before[del as usize]];
+        if i > 0 {
+          self.orig.push('\n');
+        }
+        self.orig.push_str(&fmt_rem_text_highlight(s));
+      }
+
+      // collect added lines for this hunk
+      for (i, ins) in hunk.after.enumerate() {
+        let s = self.input.interner[self.input.after[ins as usize]];
+        if i > 0 {
+          self.edit.push('\n');
+        }
+        self.edit.push_str(&fmt_add_text_highlight(s));
+      }
+    }
+
+    self.flush_changes();
+    self.output
+  }

This keeps the current no-context behavior but restores one block per hunk with correct absolute line numbers for each change region.

libs/resolver/display.rs (1)

259-315: Add a test covering multiple separated change regions

The current tests validate single-hunk cases and line-ending normalization, but don’t exercise a file with two or more non-adjacent change regions. That’s exactly where the current handle_diff implementation misbehaves (merged hunks and wrong line numbers).

Consider adding a test that changes, say, lines 5 and 20 in a 20+ line file and asserts that the diff output shows two separate blocks with the expected line numbers.

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro