From 4ee53acad0864cda11df54108ecbbc34693e8e1e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Teemu=20P=C3=A4tsi?=
 <44954973+frendsick@users.noreply.github.com>
Date: Sun, 4 May 2025 18:35:18 +0300
Subject: [PATCH] expr: Fix parsing negated character classes "[^a]" (#7884)

* expr: Fix regex escape logic

We have to track if the previous character was already escaped to determine
if the '\' character should be interpreted as an escape character.

* expr: Fix parsing caret (^) as character class negation token

* expr: Add tests for parsing carets in regex

* expr: Add missing semicolon

* expr: Simplify boolean assignment

Co-authored-by: Daniel Hofstetter <daniel.hofstetter@42dh.com>

---------

Co-authored-by: Daniel Hofstetter <daniel.hofstetter@42dh.com>
---
 src/uu/expr/src/syntax_tree.rs | 10 ++++++++--
 tests/by-util/test_expr.rs     | 24 ++++++++++++++++++++++++
 2 files changed, 32 insertions(+), 2 deletions(-)

diff --git a/src/uu/expr/src/syntax_tree.rs b/src/uu/expr/src/syntax_tree.rs
index c555e6ccc..3026d5d41 100644
--- a/src/uu/expr/src/syntax_tree.rs
+++ b/src/uu/expr/src/syntax_tree.rs
@@ -166,13 +166,19 @@ impl StringOp {
                 };
 
                 // Handle the rest of the input pattern.
-                // Escape characters that should be handled literally within the pattern.
+                // Escaped previous character should not affect the current.
                 let mut prev = first.unwrap_or_default();
+                let mut prev_is_escaped = false;
                 for curr in pattern_chars {
                     match curr {
-                        '^' if prev != '\\' => re_string.push_str(r"\^"),
+                        // Carets are interpreted literally, unless used as character class negation "[^a]"
+                        '^' if prev_is_escaped || !matches!(prev, '\\' | '[') => {
+                            re_string.push_str(r"\^");
+                        }
                         char => re_string.push(char),
                     }
+
+                    prev_is_escaped = prev == '\\' && !prev_is_escaped;
                     prev = curr;
                 }
 
diff --git a/tests/by-util/test_expr.rs b/tests/by-util/test_expr.rs
index 5ac5c262d..193737d10 100644
--- a/tests/by-util/test_expr.rs
+++ b/tests/by-util/test_expr.rs
@@ -302,6 +302,26 @@ fn test_regex() {
         .args(&["^^^^^^^^^", ":", "^^^"])
         .succeeds()
         .stdout_only("2\n");
+    new_ucmd!()
+        .args(&["ab[^c]", ":", "ab[^c]"])
+        .succeeds()
+        .stdout_only("3\n"); // Matches "ab["
+    new_ucmd!()
+        .args(&["ab[^c]", ":", "ab\\[^c]"])
+        .succeeds()
+        .stdout_only("6\n");
+    new_ucmd!()
+        .args(&["[^a]", ":", "\\[^a]"])
+        .succeeds()
+        .stdout_only("4\n");
+    new_ucmd!()
+        .args(&["\\a", ":", "\\\\[^^]"])
+        .succeeds()
+        .stdout_only("2\n");
+    new_ucmd!()
+        .args(&["^a", ":", "^^[^^]"])
+        .succeeds()
+        .stdout_only("2\n");
     new_ucmd!()
         .args(&["-5", ":", "-\\{0,1\\}[0-9]*$"])
         .succeeds()
@@ -319,6 +339,10 @@ fn test_regex() {
         .args(&["^abc", ":", "^abc"])
         .fails()
         .stdout_only("0\n");
+    new_ucmd!()
+        .args(&["abc", ":", "ab[^c]"])
+        .fails()
+        .stdout_only("0\n");
 }
 
 #[test]