mb_strimwidth inserts error markers in invalid input string (for backwards compatibility)

The old implementation did this. It also did the same to the
trim marker, if the trim marker was invalid in the specified
encoding, but I have not imitated that behavior (for performance).

Author: alexdowad

ext/mbstring/mbstring.c

@@ -2414,7 +2414,7 @@ static zend_string* mb_trim_string(zend_string *input, zend_string *marker, cons
 	int remaining_width = width;
 	unsigned int to_skip = from;
 	size_t out_len = 0;
-	bool first_call = true;
+	bool first_call = true, input_err = false;
 	mb_convert_buf buf;
 
 	while (in_len) {
@@ -2425,7 +2425,9 @@ static zend_string* mb_trim_string(zend_string *input, zend_string *marker, cons
 			to_skip -= out_len;
 		} else {
 			for (int i = to_skip; i < out_len; i++) {
-				remaining_width -= character_width(wchar_buf[i]);
+				uint32_t w = wchar_buf[i];
+				input_err |= (w == MBFL_BAD_INPUT);
+				remaining_width -= character_width(w);
 				if (remaining_width < 0) {
 					/* We need to truncate string and append trim marker */
 					width -= mb_get_strwidth(marker, enc);
@@ -2449,8 +2451,11 @@ static zend_string* mb_trim_string(zend_string *input, zend_string *marker, cons
 		first_call = false;
 	}
 
-	/* The input string is fine; we don't need to append the trim marker */
-	if (from == 0) {
+	/* The input string fits in the requested width; we don't need to append the trim marker
+	 * However, if the string contains erroneous byte sequences, those should be converted
+	 * to error markers */
+	if (from == 0 && !input_err) {
+		/* This just increments the string's refcount; it doesn't really 'copy' it */
 		return zend_string_copy(input);
 	}
 	return mb_drop_chars(input, enc, from);

ext/mbstring/tests/mb_strimwidth.phpt

@@ -192,8 +192,7 @@ testStrimwidth("\x00\x00\x00\x00+\x00\x00\x00\x00k\x00'\x11Yz", 1, 0, 'greek');
 // sometimes not skip over characters at beginning of string when requested to
 testStrimwidthWithMarker(str_repeat("a", 268), '', 12, 255, 'ASCII');
 
-// Try invalid string; as long as it fits in the specified width, where each
-// invalid character is counted as having width 1, pass it through
+// Try invalid string; invalid sequences will be converted to error markers
 testStrimwidth("\x80\x80\x80", 0, 10, 'UTF-8');
 
 // Try invalid marker
@@ -254,7 +253,7 @@ mb_strimwidth(): Argument #3 ($width) is out of range
 == Other ==
 start=1 width=0 result=[2e 2e 2e] length=3 width=3
 start=12 width=255 result=[61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61 61] length=255 width=255
-start=0 width=10 result=[80 80 80] length=3 width=3
+start=0 width=10 result=[3f 3f 3f] length=3 width=3
 start=0 width=10 result=[61 62 63 64 65 66 67 80 80 80] length=10 width=10
 start=0 width=1 result=[61 62 63 64 65 66 67 68 69 6a 6b 6c 6d 6e 6f 70] length=16 width=16
 start=-3 width=10 result=[61 62 63] length=3 width=3