What exactly IS in that file? It's impossible to test this without the data..

I generate data with simple PHP script:
<?php
file_put_contents('data', '');
for ($i = 0; $i < 1024; $i++){
    file_put_contents('data', chr(rand(0, 255)), FILE_APPEND);
}

However in php-cli substr work correctly.

If it works in CLI but not in webserver, then there's propably something different between those two, maybe PHP version..?*

The version and build are the same.

How do you know that the md5 sum for that part should be exactly that?
And we can't really test this reliably with some random data..

> How do you know that the md5 sum for that part should be exactly that?
I ran script on other PHP version and other OSes

PHP Script source: http://test.sergey89.net/test.php.txt
File containing data: http://test.sergey89.net/data
Result: http://test.sergey89.net/test.php

We can give you ftp/ssh access to test yourself.

Can you just provide the script and data in e.g. zip package somewhere where anyone can download it from? (those urls above are not working)

And shell access is not necessary for simple thing like this..especially when I'm 99% sure you're just doing something wrong.. :)

No feedback was provided for this bug for over a week, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

I have something like that on PHP 5.2.6 on Linux.
Here is a test script file:
===========================

<?php

echo "OS is ".PHP_OS."<br />\n";

echo "PHP is ".phpversion()."<br />\n";

echo "function overload = ".ini_get('mbstring.func_overload')."<br />\n";

//mb_internal_encoding('UTF-8');

echo "MB_INTERNAL_ENCODING =".mb_internal_encoding()."<br />\n";

define('IDENTIFIER_OLE', pack("CCCCCCCC",0xd0,0xcf,0x11,0xe0,0xa1,0xb1,0x1a,0xe1));

$data = file_get_contents($_SERVER['DOCUMENT_ROOT'].'/substr_bug/empty file.xls');

echo "Data length = ".strlen($data)."<br />\n";
echo "First 8 symbols  ==>".var_export(substr($data,0,8),1)."<== <br />\n";
echo "Compare result (substr(\$data,0,8)==IDENTIFIER_OLE) - ".var_export(substr($data,0,8)==IDENTIFIER_OLE,1)."<br />\n";
echo "Substring length (substr(\$data,0,8)) - ".strlen(substr($data,0,8))."<br />\n";

?>

Output:
=======
OS is Linux
PHP is 5.2.6
function overload = 0
MB_INTERNAL_ENCODING =ISO-8859-1
Data length = 13824
First 8 symbols ==>'&#1087;&#1086;&#1070;&#9553;&#9568;&#1040;'<==
Compare result (substr($data,0,8)==IDENTIFIER_OLE) - true
Substring length (substr($data,0,8)) - 8

But if you uncomment line with 
mb_internal_encoding('UTF-8');
output will be changed like that (look at the file size, result of substr and length of substr result)...

Output with mb_internal_encoding=UTF-8:
=======================================
OS is Linux
PHP is 5.2.6
function overload = 0
MB_INTERNAL_ENCODING =UTF-8
Data length = 13824
First 8 symbols ==>'&#1087;&#1086;&#1070;&#9553;&#9568;&#1040;' . "\0" . '' . "\0" . '' . "\0" . '' . "\0" . '' . "\0" . ''<==
Compare result (substr($data,0,8)==IDENTIFIER_OLE) - false
Substring length (substr($data,0,8)) - 13


mbstring.func_overload is set to 0 in .htaccess file in current dir.

"empty file.xls" is an empty MS Excel 2003 file. Can be downloaded from http://an-best.ru/empty_file.xls