Memory leak in xml extension with zend allocator

[YuanchengJiang]

Description

The following code:

<?php
function createParser(bool $huge) {
    $parser = xml_parser_create();
    xml_parser_set_option($parser, XML_OPTION_PARSE_HUGE, $huge);
    xml_set_element_handler($parser, function($parser, $data) {
    }, function($parser, $data) {
    });
    return $parser;
}
// Construct XML that is too large to parse without XML_OPTION_PARSE_HUGE
$long_text = str_repeat("A", 1000 * 1000 * 5 /* 5 MB */);
$long_xml_head = "<?xml version=\"1.0\"?><container><$long_text/><$long_text/><second>foo</second>";
$long_xml_tail = "</container>";
$parser = createParser(false);
$ret = xml_parse($parser, $long_xml_head, true);
echo "ret = $ret (", xml_error_string(xml_get_error_code($parser)), ")\n";
$parser = createParser(true);
$ret = xml_parse($parser, $long_xml_head, false);
$parser = createParser(true);
$ret = xml_parse_into_struct($parser, $long_xml_head . $long_xml_tail, $values, $index);
?>

Resulted in this output:

Fatal error: Allowed memory size of 20971520 bytes exhausted at /php-src/Zend/zend_string.h:176 (tried to allocate 5000032 bytes) in test.php on line 18

=================================================================
==4013058==ERROR: LeakSanitizer: detected memory leaks

Direct leak of 5000001 byte(s) in 1 object(s) allocated from:
    #0 0x56408aa88c1e in malloc (/php-src/sapi/cli/php+0x1c88c1e) (BuildId: 5b92f8b1df62e3423fa6490d16c9978bd9086692)
    #1 0x7f149e2c97f2 in xmlStrdup (/lib/x86_64-linux-gnu/libxml2.so.2+0xe37f2) (BuildId: aebf8e42966c3ce475ff9d9d51a762831adcbb61)

SUMMARY: AddressSanitizer: 5000001 byte(s) leaked in 1 allocation(s).

To reproduce:

USE_ZEND_ALLOC=1 php -d "memory_limit=20M" ./test.php

PHP Version

nightly

Operating System

ubuntu 22.04

[devnexen]

Reproduced even w/o USE_ZEND_ALLOC

[nielsdos]

This is actually a duplicate of #14044
While this specific instance can be fixed by removing the needless copy, it's the same fundamental issue as above.
Still, David's patch is a nice optimization.