diff: cut the -p/-F function line at 40 bytes and trim vertical tabs

src/utils.rs

@@ -182,7 +182,7 @@ pub struct FunctionLineFinder<'a> {
 }
 
 impl<'a> FunctionLineFinder<'a> {
-    /// Maximum number of characters of a function line that are shown.
+    /// Maximum number of bytes of a function line that are shown.
     const MAX_WIDTH: usize = 40;
 
     /// Returns `None` when there are no patterns, so no function lines.
@@ -209,15 +209,20 @@ impl<'a> FunctionLineFinder<'a> {
             }
         }
         self.next_line = self.next_line.max(end);
-        let line = self.last_match?.trim_ascii_start();
-        let end = match std::str::from_utf8(line) {
-            Ok(text) => text
-                .char_indices()
-                .nth(Self::MAX_WIDTH)
-                .map_or(line.len(), |(end, _)| end),
-            Err(_) => line.len().min(Self::MAX_WIDTH),
-        };
-        Some(line[..end].trim_ascii_end().to_vec())
+        // C's isspace(): unlike `is_ascii_whitespace`, it includes vertical tab.
+        let is_space = |byte: &u8| byte.is_ascii_whitespace() || *byte == b'\x0b';
+        let line = self.last_match?;
+        let start = line
+            .iter()
+            .position(|byte| !is_space(byte))
+            .unwrap_or(line.len());
+        let line = &line[start..];
+        let line = &line[..line.len().min(Self::MAX_WIDTH)];
+        let end = line
+            .iter()
+            .rposition(|byte| !is_space(byte))
+            .map_or(0, |last| last + 1);
+        Some(line[..end].to_vec())
     }
 }
 

tests/by-util/test_diff.rs

@@ -873,6 +873,47 @@ fn function_lines() {
         .stderr_is("'\\(': Unmatched ( or \\(\n");
 }
 
+// A function line is cut at 40 bytes, even inside a multi-byte character.
+#[test]
+fn function_line_is_cut_at_40_bytes() {
+    let (at, ts) = at_and_ts!();
+    let function_line = format!("func_{}()", "é".repeat(40));
+    at.write("old", &format!("{function_line}\n1\n2\n3\n4\n5\n6\n"));
+    at.write("new", &format!("{function_line}\n1\n2\n3\n4\nX\n6\n"));
+
+    let mut expected = b"@@ -3,5 +3,5 @@ ".to_vec();
+    expected.extend(&function_line.as_bytes()[..40]);
+    expected.push(b'\n');
+
+    let result = ts
+        .ucmd()
+        .args(&["-u", "-p", "old", "new"])
+        .fails_with_code(1);
+    let header = result
+        .stdout()
+        .split_inclusive(|&byte| byte == b'\n')
+        .nth(2);
+    assert_eq!(header, Some(expected.as_slice()));
+}
+
+// Vertical tabs count as white space around a function line, as they do for
+// C's isspace().
+#[test]
+fn function_line_vertical_tabs_are_trimmed() {
+    let (at, ts) = at_and_ts!();
+    at.write("old", "\x0b\x0c fg() \t\x0c\r\x0b\n1\n2\n3\n4\n5\n6\n");
+    at.write("new", "\x0b\x0c fg() \t\x0c\r\x0b\n1\n2\n3\n4\nX\n6\n");
+
+    ts.ucmd()
+        .args(&["-u", "-F", "fg", "old", "new"])
+        .fails_with_code(1)
+        .stdout_contains("\n@@ -3,5 +3,5 @@ fg()\n");
+    ts.ucmd()
+        .args(&["-c", "-F", "fg", "old", "new"])
+        .fails_with_code(1)
+        .stdout_contains("\n*************** fg()\n");
+}
+
 #[cfg(unix)]
 #[test]
 fn quoted_file_names() {