#40894 [Opn->Asn]: fix for bug #38770 causes more harm than good on big-endian 64bit

ID: 40894 Updated by: tony2001@php.net Reported By: mmarek at suse dot cz -Status: Open +Status: Assigned Bug Type: *Programming Data Structures PHP Version: 5.2.1 -Assigned To: +Assigned To: iliaa Previous Comments: ------------------------------------------------------------------------ [2007-03-22 15:59:03] mmarek at suse dot cz Description: ------------ Fix for bug #38770 breaks unpack() on big-endian 64bit, because it reads uninitialized memory or memory that doesn't relate to the data being processed. In http://cvs.php.net/viewvc.cgi/php-src/ext/standard/pack.c?r1=1.62&r2=1.63 input[inputpos + machine_endian_long_map[3]] will read input[inputpos+7] when unpacking 4 bytes. Please revert the fix, as it tries to fix a questionable issue (-30000 simply doesn't fit into unsigned type) and causes trouble in cases where it worked before (eg. pear is not even installable on POWER, b/c install-pear-nozlib.phar makes heavy use of unpack()). Reproduce code: --------------- run print_r( unpack( "V", pack( "V", 200 ) )); several times on ppc64. Expected result: ---------------- It should output Array ( [1] => 200 ) all the time. Actual result: -------------- # ./sapi/cli/php -r 'print_r( unpack( "V", pack( "V", 200 ) ));' Array ( [1] => 200 ) # ./sapi/cli/php -r 'print_r( unpack( "V", pack( "V", 200 ) ));' Array ( [1] => -2147483448 ) # ./sapi/cli/php -r 'print_r( unpack( "V", pack( "V", 200 ) ));' Array ( [1] => 200 ) # ./sapi/cli/php -r 'print_r( unpack( "V", pack( "V", 200 ) ));' Array ( [1] => 200 ) # ./sapi/cli/php -r 'print_r( unpack( "V", pack( "V", 200 ) ));' Array ( [1] => -2147483448 ) ------------------------------------------------------------------------ -- Edit this bug report at http://bugs.php.net/?id=40894&edit=1