Fixed redirection bug

[IJMacD]

In the case of MOD_REWRITE being used on an Apache server to forge the webtrees url,
webtrees would try to redirect to its source directory path and would not respect
the SERVER_URL site setting (causing an infinite redirect loop).

Further explanation:

Suppose the webtrees source is in a folder named /webtrees in the document root of the webserver but suppose you want the URL for your site to be something like http://example.com/familytree.

Previously webtrees would not handle this situation correctly and set the WT_SCRIPT_PATH constant based on its file path (/webtrees from $_SERVER['SCRIPT_NAME']). This constant however is used in URL generation e.g. LOGIN_URL. It should therefore be calculated from the incoming URL.

The result is all links generated which are based on WT_SCRIPT_PATH will take the user to a different URL root; moreover trying to set the SERVER_URL site setting in the database causes in infinite redirect loop since Apache tries to redirect to the favoured URL and webtrees sees it does not match the file directory and tries to redirect back.

This commit fixes the bug - if only for the Apache redirect case. I believe this is at least better than no fix at all.

[fisharebest]

The ZF library has functions to extract the request URL from the server environment. Would it be better to use that?

Perhaps something along these lines...

$request = new Zend_Controller_Request_Http;
echo $request->getScheme() . '://' . $request->getHttpHost() . $request->getBaseUrl();

Does this code pick up your redirection OK?

[IJMacD]

Unfortunately, it doesn't. I've looked through all the Zend documentation about the Request classes. It seems that they aren't designed to catch this case. They use the same server variables as Webtrees currently uses.

The problem is that Zend applications are designed to have a single entry point index.php for example and you would have a rewrite rule such as (the recommended):

RewriteCond %{REQUEST_FILENAME} !-f
RewriteRule ^.*$ index.php [NC,L]

This means unless the file actually exists (images, css etc.) route everything through the file index.php

Because Webtrees does not work this way and has multiple entry points e.g. index.php, login.php, individual.php the code from the Zend class can't correctly guess the base URL - it assumes you want /path/index.php instead of /path.

For reference here's the relevant part of my .htaccess file:

RewriteRule ^family$ /family/ [R=301,NC,L]
RewriteCond %{REQUEST_FILENAME} !-f
RewriteRule ^family/(.*)$ /webtrees/$1 [L]

[fisharebest]

The problem is that Zend applications are designed to have a single entry point

We know the current script name, so it is straightforward to strip that off. The Zend library Zend_Controller_Request_http::setRequestUri() appears to read the original request (whereas the WT code reads the actual file/script), so I was hoping that it would detect your request (family) rather than the actual path (webtrees).

FYI, the intention is to move to ZF2 and a single entry point (i.e. a front controller).

I don't have apache on my dev machine (I prefer nginx), so I'll need to set it up to investigate...

[fisharebest]

Thanks!