Propagate UTF-8 flag during Rope operations (php#10915)

Author: Girgias

Zend/zend_vm_def.h

@@ -3330,8 +3330,6 @@ ZEND_VM_HANDLER(56, ZEND_ROPE_END, TMP, CONST|TMPVAR|CV, NUM)
 	zend_string **rope;
 	zval *var, *ret;
 	uint32_t i;
-	size_t len = 0;
-	char *target;
 
 	rope = (zend_string**)EX_VAR(opline->op1.var);
 	if (OP2_TYPE == IS_CONST) {
@@ -3364,12 +3362,18 @@ ZEND_VM_HANDLER(56, ZEND_ROPE_END, TMP, CONST|TMPVAR|CV, NUM)
 			}
 		}
 	}
+
+	size_t len = 0;
+	uint32_t flags = ZSTR_COPYABLE_CONCAT_PROPERTIES;
 	for (i = 0; i <= opline->extended_value; i++) {
+		flags &= ZSTR_GET_COPYABLE_CONCAT_PROPERTIES(rope[i]);
 		len += ZSTR_LEN(rope[i]);
 	}
 	ret = EX_VAR(opline->result.var);
 	ZVAL_STR(ret, zend_string_alloc(len, 0));
-	target = Z_STRVAL_P(ret);
+	GC_ADD_FLAGS(Z_STR_P(ret), flags);
+
+	char *target = Z_STRVAL_P(ret);
 	for (i = 0; i <= opline->extended_value; i++) {
 		memcpy(target, ZSTR_VAL(rope[i]), ZSTR_LEN(rope[i]));
 		target += ZSTR_LEN(rope[i]);

Zend/zend_vm_execute.h

@@ -3104,13 +3104,16 @@ static zend_string* ZEND_FASTCALL zend_jit_rope_end(zend_string **rope, uint32_t
 	zend_string *ret;
 	uint32_t i;
 	size_t len = 0;
-	char *target;
 
+	uint32_t flags = ZSTR_COPYABLE_CONCAT_PROPERTIES;
 	for (i = 0; i <= count; i++) {
+		flags &= ZSTR_GET_COPYABLE_CONCAT_PROPERTIES(rope[i]);
 		len += ZSTR_LEN(rope[i]);
 	}
 	ret = zend_string_alloc(len, 0);
-	target = ZSTR_VAL(ret);
+	GC_ADD_FLAGS(ret, flags);
+
+	char *target = ZSTR_VAL(ret);
 	for (i = 0; i <= count; i++) {
 		memcpy(target, ZSTR_VAL(rope[i]), ZSTR_LEN(rope[i]));
 		target += ZSTR_LEN(rope[i]);

ext/opcache/jit/zend_jit_helpers.c

@@ -107,6 +107,13 @@ $s = $o . $o;
 var_dump($s);
 var_dump(zend_test_is_string_marked_as_valid_utf8($s));
 
+echo "Rope concat:\n";
+$foo = 'f';
+$bar = 'b';
+$baz = 'a';
+$rope = "$foo$bar$baz";
+var_dump(zend_test_is_string_marked_as_valid_utf8($rope));
+
 echo "str_repeat:\n";
 $string = "a";
 $string_concat = str_repeat($string, 100);
@@ -183,6 +190,8 @@ bool(true)
 Concatenation of objects:
 string(2) "zz"
 bool(true)
+Rope concat:
+bool(true)
 str_repeat:
 bool(true)
 bool(false)

ext/zend_test/tests/strings_marked_as_utf8.phpt

@@ -102,6 +102,12 @@ $s = $o . $o;
 $s = $s . $non_utf8;
 var_dump(zend_test_is_string_marked_as_valid_utf8($s));
 
+echo "Rope concat:\n";
+$foo = 'f';
+$bar = "\xc3";
+$baz = 'a';
+$rope = "$foo$bar$baz";
+var_dump(zend_test_is_string_marked_as_valid_utf8($rope));
 ?>
 --EXPECT--
 Integer cast to string concatenated to invalid UTF-8:
@@ -129,3 +135,5 @@ Concatenation in loop (compound assignment):
 bool(false)
 Concatenation of objects:
 bool(false)
+Rope concat:
+bool(false)