From 2712cbf747287470c17c9a39512b3a1f51dc7ed3 Mon Sep 17 00:00:00 2001
From: Jayesh <jayesh.bhade@keychain.com>
Date: Tue, 4 Aug 2026 19:28:48 +0530
Subject: [PATCH] fix: keep the em/strong mask the same length as the source

emStrong and del align maskedSrc with src by slicing the mask from the
end, so every mask must keep the length of the text it replaces. The
escape mask substituted a fixed two-character '++', but anyPunctuation is
/\\([\p{P}\p{S}])/gu and matches astral code points, so an escaped
character above U+FFFF is three code units masked as two.

The mask then ended up shorter than the source, the clip reached one code
unit too far back, and the resulting match.index was applied to src: the
closing delimiter was consumed a code unit late, leaving a stray
delimiter inside the emphasis and, for some inputs, dropping a character
of the document.

Substitute one '+' per code unit instead. '+' is punctuation, so the
delimiter flanking of the masked region is unchanged, and the ASCII case
still produces '++'.
---
 src/Lexer.ts                                     |  7 +++++--
 .../specs/new/em_escaped_astral_punctuation.html |  7 +++++++
 test/specs/new/em_escaped_astral_punctuation.md  | 16 ++++++++++++++++
 3 files changed, 28 insertions(+), 2 deletions(-)
 create mode 100644 test/specs/new/em_escaped_astral_punctuation.html
 create mode 100644 test/specs/new/em_escaped_astral_punctuation.md

diff --git a/src/Lexer.ts b/src/Lexer.ts
index 89e291f19c..64cbe8101c 100644
--- a/src/Lexer.ts
+++ b/src/Lexer.ts
@@ -316,8 +316,11 @@ export class _Lexer<ParserOutput = string, RendererOutput = string> {
       }
     }
 
-    // Mask out escaped characters
-    maskedSrc = maskedSrc.replace(this.tokenizer.rules.inline.anyPunctuation, '++');
+    // Mask out escaped characters.
+    // Every mask must keep the length it replaces: emStrong and del line
+    // maskedSrc up with src by slicing from the end. `anyPunctuation` matches
+    // unicode punctuation, so an escaped astral character is 3 code units.
+    maskedSrc = maskedSrc.replace(this.tokenizer.rules.inline.anyPunctuation, match0 => '+'.repeat(match0.length));
 
     // Mask out other blocks
     maskedSrc = maskedSrc.replace(this.tokenizer.rules.inline.blockSkip, (match0, _link, context) => {
diff --git a/test/specs/new/em_escaped_astral_punctuation.html b/test/specs/new/em_escaped_astral_punctuation.html
new file mode 100644
index 0000000000..588d3bab26
--- /dev/null
+++ b/test/specs/new/em_escaped_astral_punctuation.html
@@ -0,0 +1,7 @@
+<p><em>a</em>\🙂</p>
+<p><em>a</em> \🙂 <em>b</em></p>
+<p>x<em>ab</em>\🙂y</p>
+<p><em>a</em>a\𐄀<em>b</em></p>
+<p><strong>a</strong>\🙂b</p>
+<p><del>a</del>\🙂</p>
+<p><em>foo\🙂bar</em></p>
diff --git a/test/specs/new/em_escaped_astral_punctuation.md b/test/specs/new/em_escaped_astral_punctuation.md
new file mode 100644
index 0000000000..541830ca73
--- /dev/null
+++ b/test/specs/new/em_escaped_astral_punctuation.md
@@ -0,0 +1,16 @@
+---
+gfm: true
+---
+*a*\🙂
+
+*a* \🙂 *b*
+
+x*ab*\🙂y
+
+*a*a\𐄀*b*
+
+__a__\🙂b
+
+~~a~~\🙂
+
+*foo\🙂bar*