CiviCRM Community Forums (archive)

We run CiviCRM on drupal 7, on an intranet server. Most of the time, most people access the server using just the server name, e.g. http://lcbru-test/ and this works fine, as does the canonical server name of http://lcbru-test.xuhl-tr.nhs.uk/

However, since upgrading to 4.4.0 we've discovered a problem - when accessing the server using the short version, the resource URLs formed internally to CiviCRM to bring in javascript and css do not get formed correctly. This was not a problem before 4.4.0 so I assume there is a change in the way these URLs are being constructed.

What we now see is:

<script type="text/javascript" src="http://lcbru-test/.xuhl-tr.nhs.uk/sites/all/modules/civicrm/packages/jquery/jquery-1.8.3.min.js?mvjnlx"></script>
Something is adding the domain to the server name, but only after a closing slash. Understandably, the web browsers do not process this correctly.

I've tried to track down where in the CiviCRM code this URL is built but I'm afraid I got lost in CRM/Core/Resources.php and the methods therein.

As a workaround, I've configured apache to redirect requests from the short form to the long form, and that works fine. But it might be worth checking out why the URL isn't being built correctly - ours might not be the only situation in which the URL is malformed.

Hi Xavier,

The CiviCRM Resource URL in that admin page is set correctly, even when accessing the site using the short form of the server name.

It says "http://lcbru-test.xuhl-tr.nhs.uk/sites/all/modules/civicrm"

My civicrm.settings.php has:
define( 'CIVICRM_UF_BASEURL'      , 'http://lcbru-test.xuhl-tr.nhs.uk/' );

So it appears that in 4.4.0 this Resource URL isn't being used for bringing in the jquery scripts, but instead something is being generated on the fly based on the server name and the domain being botched together.

It's complicated.

Our live system was updated all the way from 4.2.4 to 4.4.0 on Thursday, and it was the live system in which we noticed the behaviour first. We'd had no reports of the problem from people using the test system, which had been upgraded to 4.3.5 some weeks ago. This might suggest the problem was introduced later than 4.3.5, but it could just be that no-one had cause to use the test system via the short URL in the intervening weeks. Sorry, that's not very helpful. All I can say for certain is that I can't rule out a 4.3.x change being at fault.