Description of problem: I am usin a new RH EL 4 and have a problem with this function. The output is not corect and I tryed it also on some other systems and they also have this same bug. On my old server with RH 7.3 is answer corect. I think that is architecture independent, because do it also on i386 systems. There is some informations about system : affected system: Red Hat Enterprise Linux ES release 4 (Nahant Update 4) Linux www.prolux.cz 2.6.9-42.0.3.ELsmp #1 SMP Mon Sep 25 17:24:31 EDT 2006 x86_64 x86_64 x86_64 GNU/Linux php-4.3.9-3.18 correct working system: Red Hat Linux release 7.3 (Valhalla) Linux prolux.speednet.cz 2.4.20-28.7 #1 Thu Dec 18 11:23:36 EST 2003 i686 unknown php-4.1.2-7.3.6 sorce code of php function : <? function L($x, $n) { return ( ($x<<$n) | ((pow(2, $n) - 1) & ($x>>(32-$n))) ); } ?> <html> <head> <meta http-equiv="Content-Type" content="text/html; charset=windows-1250"> <title></title> </head> <body> <p><font size="4">Test of function: return ( ($x<<$n) | ((pow(2, $n) - 1) & ($x>>(32-$n))) );</font></p> <table border="0" cellspacing="1" width="600"> <tr> <td colspan="2">This is the correct answer:</td> </tr> <tr> <td><font color="#808080">INPUT: </font>-3086048340</td> <td><font color="#808080">OUTPUT:</font> 30700937</td> </tr> <tr> <td colspan="2"><hr noshade></td> </tr> <tr> <td colspan="2">Aswer of this function on this server:</td> </tr> <tr> <td><font color="#808080">input: </font>-3086048340</td> <td><font color="#808080">output:</font> <? echo L(-3086048340, 5) ?></td> </tr> </table> </body> </html> How reproducible: Do it all times on affected server Steps to Reproduce: 1. page on affected server with this source code : http://www2.prolux.cz/tmp/test.php 2. page on corect working server with same souce php code : http://router.prolux.cz/test.php Thank you very much for answer, With the best regards Vasek Valenta
Thanks for the report. The result of "-3086048340 << 5" will overflow signed 32-bit integer arithemetic (as used internally by PHP) so the output of that function will not be well-defined for that input. If you need to reliably manipulate large integers then the "gmp" extension can be used instead.