ptx: fix panic when truncation string/keyword contain multibyte Unicode (#10836)

Co-authored-by: Sylvestre Ledru <sylvestre@debian.org>

Author: Xylphy

src/uu/ptx/src/ptx.rs

@@ -459,11 +459,11 @@ fn get_output_chunks(
     // https://github.com/MaiZure/coreutils-8.3/blob/master/src/ptx.c#L1234
     let half_line_size = config.line_width / 2;
     let max_before_size = cmp::max(half_line_size as isize - config.gap_size as isize, 0) as usize;
+
+    let keyword_len = keyword.chars().count();
+    let trunc_len = config.trunc_str.chars().count();
     let max_after_size = cmp::max(
-        half_line_size as isize
-            - (2 * config.trunc_str.len()) as isize
-            - keyword.len() as isize
-            - 1,
+        half_line_size as isize - (2 * trunc_len) as isize - keyword_len as isize - 1,
         0,
     ) as usize;
 
@@ -504,7 +504,7 @@ fn get_output_chunks(
     // and get the string
     let after_str: String = all_after[0..after_end].iter().collect();
     after.push_str(&after_str);
-    assert!(max_after_size >= after.len());
+    assert!(max_after_size >= after.chars().count());
 
     // the tail chunk
 

tests/by-util/test_ptx.rs

@@ -339,6 +339,18 @@ fn test_unicode_truncation_alignment() {
         .stdout_only("     /   bar\n        föö/\n");
 }
 
+#[test]
+fn test_unicode_in_after_chunk_does_not_panic() {
+    // Regression test for a panic in get_output_chunks() when the computed
+    // max_after_size used byte lengths but the output was assembled as chars.
+    // The emoji is multibyte in UTF-8 and previously could trigger:
+    // `assertion failed: max_after_size >= after.len()`.
+    new_ucmd!()
+        .pipe_in("We've got +11 more G of 1.70. 🛠\n")
+        .succeeds()
+        .stdout_contains("We've got +11");
+}
+
 #[test]
 fn test_duplicate_input_files() {
     new_ucmd!()