From d378aedd3c2a811c2baa56d070b77418e4abe0e4 Mon Sep 17 00:00:00 2001
From: Rahul Ranjan <rahulsingh.ranjan06@gmail.com>
Date: Fri, 13 Nov 2026 07:12:03 +0000
Subject: [PATCH] fix: don't read past position 0 when checking dollar escaping
 in dollarmath

is_escaped decremented its position index inside a loop whose condition
allowed back_pos == 0, so on Python's negative indexing it read the last
character of the inline source instead of finding no preceding
character. A dollar at position 0 was then treated as escaped whenever
the inline source ended with a backslash, silently dropping math at the
start of a paragraph, and backslash runs before a mid-source dollar
picked up trailing characters, flipping odd counts to even.

Co-authored-by: rahul05ranjan <11793455+rahul05ranjan@users.noreply.github.com>
---
 mdit_py_plugins/dollarmath/index.py |  2 +-
 tests/test_dollarmath.py            | 35 +++++++++++++++++++++++++++++
 2 files changed, 36 insertions(+), 1 deletion(-)

diff --git a/mdit_py_plugins/dollarmath/index.py b/mdit_py_plugins/dollarmath/index.py
index c773c54..993da82 100644
--- a/mdit_py_plugins/dollarmath/index.py
+++ b/mdit_py_plugins/dollarmath/index.py
@@ -157,7 +157,7 @@ def is_escaped(state: StateInline, back_pos: int, mod: int = 0) -> bool:
     """Test if dollar is escaped."""
     # count how many \ are before the current position
     backslashes = 0
-    while back_pos >= 0:
+    while back_pos > 0:
         back_pos = back_pos - 1
         if state.src[back_pos] == "\\":
             backslashes += 1
diff --git a/tests/test_dollarmath.py b/tests/test_dollarmath.py
index 71c199e..5575f03 100644
--- a/tests/test_dollarmath.py
+++ b/tests/test_dollarmath.py
@@ -164,3 +164,38 @@ def test_label_is_html_escaped(label_normalizer, label):
         f'<a href="#{label}" class="mathlabel" title="Permalink to this equation">'
         "¶</a>\na=1\n</div>\n"
     )
+
+
+def test_inline_func_start_with_trailing_backslash():
+    """A dollar at position 0 must not read backslashes from the *end* of the source.
+
+    ``is_escaped`` used to decrement its position index past 0 and rely on
+    negative indexing, so it read the last character of the source instead of
+    finding no preceding character. Math at position 0 was then dropped
+    whenever the inline source ended with a backslash.
+    """
+    inline_func = main.math_inline_dollar()
+    md = MarkdownIt()
+    for src in ("$a$ is math\\", "$a$ is math\\\\"):
+        tokens = []
+        state = StateInline(src, md, {}, tokens)
+        assert inline_func(state, False) is True
+        assert tokens[0].type == "math_inline"
+        assert tokens[0].content == "a"
+
+
+def test_inline_func_escaped_dollar_mid_source_with_trailing_backslash():
+    """An escaped dollar mid-source must stay escaped even if the source ends with backslashes.
+
+    With the same off-the-end read as the previous test, a backslash run
+    preceding the dollar could pick up the trailing character of the source
+    and turn an odd backslash count into an even one, wrongly treating an
+    escaped dollar as the opening delimiter of a math span.
+    """
+    inline_func = main.math_inline_dollar()
+    md = MarkdownIt()
+    tokens = []
+    state = StateInline("\\$x$ y\\", md, {}, tokens)
+    state.pos = 1
+    assert inline_func(state, False) is False
+    assert tokens == []