From c2cc79daa747bbbfcdec22bf1eea95a1d1ef8aa2 Mon Sep 17 00:00:00 2001
From: lenamonj <jlenamon@gmail.com>
Date: Wed, 29 Jul 2026 15:30:07 -0400
Subject: [PATCH 1/2] Fix --sanitize passing through the bidi control
 characters U+200E, U+200F and U+061C

---
 src/preprocessor.rs | 43 +++++++++++++++++++++++++++++++++++++++----
 1 file changed, 39 insertions(+), 4 deletions(-)

diff --git a/src/preprocessor.rs b/src/preprocessor.rs
index 971ae3d4..a9b8e712 100644
--- a/src/preprocessor.rs
+++ b/src/preprocessor.rs
@@ -168,7 +168,7 @@ pub fn sanitize(line: &str) -> String {
 #[inline]
 fn is_sanitize_trigger(b: u8) -> bool {
     // C0 controls minus \t \n \f; DEL; UTF-8 leads with dangerous codepoints.
-    matches!(b, 0x00..=0x08 | 0x0B | 0x0D..=0x1F | 0x7F | 0xC2 | 0xE2 | 0xEF)
+    matches!(b, 0x00..=0x08 | 0x0B | 0x0D..=0x1F | 0x7F | 0xC2 | 0xD8 | 0xE2 | 0xEF)
 }
 
 /// Substitutes the byte/sequence at `bytes[i]` (or passes it through on
@@ -200,8 +200,14 @@ fn sanitize_at(
             buffer.push('\u{FFFD}');
             3
         }
+        // 0xD8 0x9C = U+061C (Arabic letter mark, a bidi control). The rest of
+        // the 0xD8 block is ordinary Arabic text.
+        0xD8 if bytes.get(i + 1) == Some(&0x9C) => {
+            buffer.push('\u{FFFD}');
+            2
+        }
         // False-alarm trigger: pass the full UTF-8 sequence through.
-        lead @ (0xC2 | 0xE2 | 0xEF) => {
+        lead @ (0xC2 | 0xD8 | 0xE2 | 0xEF) => {
             let n = utf8_len_from_lead(lead);
             buffer.push_str(&full[i..i + n]);
             n
@@ -217,10 +223,11 @@ fn sanitize_at(
 
 #[inline]
 fn is_dangerous_e2(bytes: &[u8], i: usize) -> bool {
-    // U+200B..D (zero-width), U+202A..E (bidi controls), U+2066..9 (isolates).
+    // U+200B..D (zero-width), U+200E..F (LRM/RLM), U+202A..E (bidi embedding
+    // and override), U+2066..9 (bidi isolates).
     matches!(
         (bytes.get(i + 1), bytes.get(i + 2)),
-        (Some(0x80), Some(0x8B..=0x8D | 0xAA..=0xAE)) | (Some(0x81), Some(0xA6..=0xA9))
+        (Some(0x80), Some(0x8B..=0x8F | 0xAA..=0xAE)) | (Some(0x81), Some(0xA6..=0xA9))
     )
 }
 
@@ -411,6 +418,34 @@ fn test_sanitize_substitutes_bidi_and_zero_width() {
     assert_eq!(sanitize("a\u{FEFF}b"), format!("a{r}b"));
 }
 
+#[test]
+fn test_sanitize_substitutes_every_bidi_control() {
+    // Unicode Bidi_Control is exactly these 12 codepoints. Covering only some
+    // of them leaves a reordering attack available through the rest.
+    let r = '\u{FFFD}';
+    for c in [
+        '\u{061C}', '\u{200E}', '\u{200F}', '\u{202A}', '\u{202B}', '\u{202C}', '\u{202D}',
+        '\u{202E}', '\u{2066}', '\u{2067}', '\u{2068}', '\u{2069}',
+    ] {
+        assert_eq!(
+            sanitize(&format!("a{c}b")),
+            format!("a{r}b"),
+            "U+{:04X} was not substituted",
+            c as u32
+        );
+    }
+}
+
+#[test]
+fn test_sanitize_preserves_arabic_sharing_the_alm_lead_byte() {
+    // U+061C is reached via lead byte 0xD8, which also leads ordinary Arabic.
+    assert_eq!(
+        sanitize("\u{0600}\u{061F}\u{06FF}"),
+        "\u{0600}\u{061F}\u{06FF}"
+    );
+    assert_eq!(sanitize("مرحبا بالعالم"), "مرحبا بالعالم");
+}
+
 #[test]
 fn test_sanitize_preserves_legitimate_bytes() {
     assert_eq!(sanitize("crlf\r\nline\r\n"), "crlf\r\nline\r\n");

From df382256532b96da7cc8eba8fb5dca2d67c5d8ce Mon Sep 17 00:00:00 2001
From: lenamonj <jlenamon@gmail.com>
Date: Wed, 29 Jul 2026 15:38:59 -0400
Subject: [PATCH 2/2] Add CHANGELOG entry for #3862

---
 CHANGELOG.md | 1 +
 1 file changed, 1 insertion(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 0cf4f8f7..7e0abc18 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -25,6 +25,7 @@
 
 ## Bugfixes
 - Fix `--list-languages` respecting `--paging=never`, see #3828 (@cyphercodes)
+- Fix `--sanitize` passing through the bidi control characters U+200E, U+200F and U+061C, see #3862 (@lenamonj)
 - `--strip-ansi`: also strip 8-bit C1 introducers (U+0090, U+0098, U+009B, U+009D, U+009E, U+009F) and DCS/SOS/PM/APC sequence bodies, which previously passed through. See #3729 (@curious-rabbit)
 - Fix `--ignored-suffix` not falling back to first-line/shebang detection when the ignored suffix is also a registered extension (e.g. `--ignored-suffix .txt` on a shebang script), see #2745 and #3816 (@adnrivera)
 - Fix `capacity overflow` panic when printing a snip separator at `--terminal-width=1` with multiple line ranges. Closes #3803, see #3804 (@leeewee)