Segmentation fault (11

[dcoonich]

Hi,
On FC4 with latest patches as of this date of this message. I added the following to my .htaccess at the root of the Sugar CRM install and it appears to work. I didn't troubleshoot as to which of these two entries fix it but since geeklog also had similar behaviour and it needed this to allow logins I tried it right away and it worked.

php_value register_globals 1
php_value register_long_arrays 1

This modified .htaccess file will also now allow the sample data to populate mysql and not hang!

In a nutshell add this your .htaccess BEFORE invoking the installer through the URL. Maybe this can be added to the checklist in the installer?

[SerSiTiv]

Did as suggested and Sugar stoped crashing my Apache.

So what exactly is happening that is causing the crashes?

[sugarchris]

The exact reason is fairly simple but not intuitive.

In load_user() or authenticate_user(), methods in the User class, there was some old cade that performed an array_merge() on $_SESSION. In PHP:

PHP Code:

$newArr = array(x, y, z ....);
$_SESSION = array_merge($_SESSION, $newArr); 


Intuitively, since $_SESSION is just an array, albeit an important and unique one, the above code should not produce errors. However, in certain versions of PHP 4 and 5, one can cause an Apache segfault by performing this operation.

The technical reason (and keep in mind that this is 3rd-hand information) is that the memory space occupied by $_SESSIONs is protected, The way array_merge() operates is to take the passed arrays, merge them in some memory space, then assign that new array to the value you want. Doing this with $_SESSION corrupts the memory space, and in some cases, fails catostrophically, killing the apache instance serving that page. This extends to all Superglobal class arrays, such as $_SERVER, $_POST, etc.

I believe Jacob, our CTO, can explain it much more precisely. But in a nutshell, don't do it!