json_encode: Escape U+2028 and U+2029 more often.

These characters are illegal in Javascript, so leaving them unescaped
is risky. The default encoder ($flags = 0) is fine, but the encoder
with JSON_UNESCAPED_UNICODE flag is not.

In case anyone wants the ability to leave these characters unescaped,
provide JSON_UNESCAPED_LINE_TERMINATORS.

Author: kohler

ext/json/json.c

@@ -117,6 +117,7 @@ static PHP_MINIT_FUNCTION(json)
 	PHP_JSON_REGISTER_CONSTANT("JSON_UNESCAPED_UNICODE", PHP_JSON_UNESCAPED_UNICODE);
 	PHP_JSON_REGISTER_CONSTANT("JSON_PARTIAL_OUTPUT_ON_ERROR", PHP_JSON_PARTIAL_OUTPUT_ON_ERROR);
 	PHP_JSON_REGISTER_CONSTANT("JSON_PRESERVE_ZERO_FRACTION", PHP_JSON_PRESERVE_ZERO_FRACTION);
+	PHP_JSON_REGISTER_CONSTANT("JSON_UNESCAPED_LINE_TERMINATORS", PHP_JSON_UNESCAPED_LINE_TERMINATORS);
 
 	/* options for json_decode */
 	PHP_JSON_REGISTER_CONSTANT("JSON_OBJECT_AS_ARRAY", PHP_JSON_OBJECT_AS_ARRAY);

ext/json/json_encoder.c

@@ -321,7 +321,7 @@ static void php_json_escape_string(smart_str *buf, char *s, size_t len, int opti
 
 	do {
 		us = (unsigned char)s[pos];
-		if (us >= 0x80 && !(options & PHP_JSON_UNESCAPED_UNICODE)) {
+		if (us >= 0x80 && (!(options & PHP_JSON_UNESCAPED_UNICODE) || us == 0xE2)) {
 			/* UTF-8 character */
 			us = php_next_utf8_char((const unsigned char *)s, len, &pos, &status);
 			if (status != SUCCESS) {
@@ -332,6 +332,15 @@ static void php_json_escape_string(smart_str *buf, char *s, size_t len, int opti
 				smart_str_appendl(buf, "null", 4);
 				return;
 			}
+			/* Escape U+2028/U+2029 line terminators, UNLESS both
+			   JSON_UNESCAPED_UNICODE and
+			   JSON_UNESCAPED_LINE_TERMINATORS were provided */
+			if ((options & PHP_JSON_UNESCAPED_UNICODE)
+				&& ((options & PHP_JSON_UNESCAPED_LINE_TERMINATORS)
+					|| us < 0x2028 || us > 0x2029)) {
+				smart_str_appendl(buf, &s[pos - 3], 3);
+				continue;
+			}
 			/* From http://en.wikipedia.org/wiki/UTF16 */
 			if (us >= 0x10000) {
 				unsigned int next_us;

ext/json/php_json.h

@@ -67,6 +67,7 @@ typedef enum {
 #define PHP_JSON_UNESCAPED_UNICODE       (1<<8)
 #define PHP_JSON_PARTIAL_OUTPUT_ON_ERROR (1<<9)
 #define PHP_JSON_PRESERVE_ZERO_FRACTION  (1<<10)
+#define PHP_JSON_UNESCAPED_LINE_TERMINATORS (1<<11)
 
 /* json_decode() options */
 #define PHP_JSON_OBJECT_AS_ARRAY         (1<<0)

ext/json/tests/json_encode_u2028_u2029.phpt

@@ -0,0 +1,36 @@
+--TEST--
+json_encode() tests for U+2028, U+2029
+--SKIPIF--
+<?php if (!extension_loaded("json")) print "skip"; ?>
+--FILE--
+<?php
+var_dump(json_encode(array("a\xC3\xA1b")));
+var_dump(json_encode(array("a\xC3\xA1b"), JSON_UNESCAPED_UNICODE));
+var_dump(json_encode("a\xE2\x80\xA7b"));
+var_dump(json_encode("a\xE2\x80\xA7b", JSON_UNESCAPED_UNICODE));
+var_dump(json_encode("a\xE2\x80\xA8b"));
+var_dump(json_encode("a\xE2\x80\xA8b", JSON_UNESCAPED_UNICODE));
+var_dump(json_encode("a\xE2\x80\xA8b", JSON_UNESCAPED_LINE_TERMINATORS));
+var_dump(json_encode("a\xE2\x80\xA8b", JSON_UNESCAPED_UNICODE | JSON_UNESCAPED_LINE_TERMINATORS));
+var_dump(json_encode("a\xE2\x80\xA9b"));
+var_dump(json_encode("a\xE2\x80\xA9b", JSON_UNESCAPED_UNICODE));
+var_dump(json_encode("a\xE2\x80\xA9b", JSON_UNESCAPED_LINE_TERMINATORS));
+var_dump(json_encode("a\xE2\x80\xA9b", JSON_UNESCAPED_UNICODE | JSON_UNESCAPED_LINE_TERMINATORS));
+var_dump(json_encode("a\xE2\x80\xAAb"));
+var_dump(json_encode("a\xE2\x80\xAAb", JSON_UNESCAPED_UNICODE));
+?>
+--EXPECT--
+string(12) "["a\u00e1b"]"
+string(8) "["aáb"]"
+string(10) ""a\u2027b""
+string(7) ""a‧b""
+string(10) ""a\u2028b""
+string(10) ""a\u2028b""
+string(10) ""a\u2028b""
+string(7) ""a
b""
+string(10) ""a\u2029b""
+string(10) ""a\u2029b""
+string(10) ""a\u2029b""
+string(7) ""a
b""
+string(10) ""a\u202ab""
+string(7) ""a‪b""