gitignore: strip UTF-8 BOM as git would

GitIgnoreBuilder::add() would do the same thing, but we use add_line() to
control file reading behavior.

Fixes #9105

Author: yuja

CHANGELOG.md

@@ -31,6 +31,8 @@ to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ### Fixed bugs
 
+* `.gitignore` with UTF-8 BOM can now be parsed correctly.
+
 ## [0.39.0] - 2026-03-04
 
 ### Release highlights

lib/src/gitignore.rs

@@ -68,7 +68,7 @@ impl GitIgnoreFile {
         input: &[u8],
     ) -> Result<Arc<Self>, GitIgnoreError> {
         let mut builder = gitignore::GitignoreBuilder::new(prefix);
-        for (i, input_line) in input.split(|b| *b == b'\n').enumerate() {
+        for (i, input_line) in strip_bom(input).split(|b| *b == b'\n').enumerate() {
             if input_line.starts_with(b"#") {
                 continue;
             }
@@ -153,6 +153,10 @@ impl GitIgnoreFile {
     }
 }
 
+fn strip_bom(text: &[u8]) -> &[u8] {
+    text.strip_prefix("\u{feff}".as_bytes()).unwrap_or(text)
+}
+
 #[cfg(test)]
 mod tests {
 
@@ -408,6 +412,12 @@ mod tests {
         assert!(matches(b"a/x**y/b\n", "a/xzzzy/b"));
     }
 
+    #[test]
+    fn test_gitignore_with_utf8_bom() {
+        assert!(matches(b"\xef\xbb\xbffoo\n", "foo"));
+        assert!(!matches(b"\n\xef\xbb\xbffoo\n", "foo"));
+    }
+
     #[test]
     fn test_gitignore_line_ordering() {
         assert!(matches(b"foo\n!foo/bar\n", "foo"));