Bug #81481 [Opn->Ver]: xml_get_current_byte_index limited to 32-bit numbers on 64-bit builds

Edit report at https://bugs.php.net/bug.php?id=81481&edit=1 ID: 81481 Updated by: cmb@php.net Reported by: dev at b65sol dot com Summary: xml_get_current_byte_index limited to 32-bit numbers on 64-bit builds -Status: Open +Status: Verified Type: Bug Package: XML Reader Operating System: 64-bit Linux PHP Version: 8.0.11 Block user comment: N Private report: N New Comment: For compatibility with expat 1, XML_GetCurrentByteIndex[1] should indeed return long[2]. Care should be taken that no wrap around occurs, since parser->parser->input->consumed is an unsigned long. Anyhow, returning long wouldn't make a difference on Windows (LLP64). The situation is different for expat 2, which returns XML_Index[3] which depends on a configuration value[4]. I'm not sure, however, whether expat 2 is supposed to be supported at all. We should also have a look at XML_GetCurrentByteCount[5], which is currently the same as XML_GetCurrentByteIndex, but is supposed to return the number of bytes in the current event[6]. [1] <https://github.com/php/php-src/blob/php-7.4.24/ext/xml/compat.c#L705-L710> [2] <https://github.com/libexpat/libexpat/blob/V1_0/expat/xmlparse/xmlparse.h#L364> [3] <https://github.com/libexpat/libexpat/blob/R_2_4_1/expat/lib/expat.h#L932> [4] <https://github.com/libexpat/libexpat/blob/R_2_4_1/expat/lib/expat_external.h#L153-L159> [5] <https://github.com/php/php-src/blob/php-7.4.24/ext/xml/compat.c#L712-L719> [6] <https://github.com/libexpat/libexpat/blob/R_2_4_1/expat/doc/reference.html#L2043-L2054> Previous Comments: ------------------------------------------------------------------------ [2021-09-26 15:41:47] dev at b65sol dot com Description: ------------ xml_get_current_byte_index is limited to 32-bit integers even on 64-bit builds, resulting in negative or incorrectly small results after parsing 2GiB or more XML text, which is well below PHP_INT_MAX. To get the "Expected" results listed, I changed the definition on XML_GetCurrentByteIndex to be long instead of int, but I'm unsure if that's a truly portable fix. Test script: --------------- <?php $parser = xml_parser_create('UTF-8'); xml_set_element_handler( $parser, 'startelement', null ); $emptylong = str_repeat(' ', 1024*1024); xml_parse($parser, '<root><i></i><b/><ext>Hello</ext>', false); for($i = 0; $i < 2896; $i++) { xml_parse($parser, $emptylong, false); } xml_parse($parser, '<ext></ext><ext></ext></root>', false); function startelement($parser, $name, $attribute) { if ( $name == 'EXT' ) { echo "Byte Index:".xml_get_current_byte_index($parser)."\n"; } } Expected result: ---------------- Byte Index:21 Byte Index:3036676133 Byte Index:3036676144 Actual result: -------------- Byte Index:21 Byte Index:-1258291163 Byte Index:-1258291152 ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=81481&edit=1