Based on patch by Nishanth Aravamudan <nish.aravamudan@canonical.com>

Fix the tests to handle the new, more permissive behaviour of SimpleSAML_Configuration::getBaseURL(), and add tests for getBasePath() too.

Make sure the new SimpleSAML_Configuration::getBasePath() complies with its phpdoc, returning always at least a slash. Fix also the regex that's supposed to catch paths, as it still wants them to end with a slash.

Add a new SimpleSAML_Configuration::getBasePath() method as described in #364, deprecate getBaseURL() and make sure it generates a warning (and only one) when used.

Change the extension of the documentation files from .txt to .md so that they can processed as markdown and displayed in github.

Update the documentation regarding translation (remove all references to the translation portal, reformat and fix typos).

Update the documentation. Trailing slashes are no longer required for 'baseurlpath'. This resolves #363.

Fix build. There was a bug in SimpleSAML\Utils\HTTP::parseQueryString() that was producing an empty parameter with an empty value when the query string was empty.

Change the implementation of SimpleSAML\Utils\HTTP::getSelfURL() and getSelfURLNoQuery() to honor the 'baseurlpath' configuration option instead of simply using the environment. They were actually broken since they were using it to build the scheme, host and port, but completely ignoring the path, rendering wrong URLs in between what was configured in 'baseurlpath' and the real information in the environment. This resolves #396, but also affects #5. The changes to getSelfURLNoQuery() in #391 are unnecessary now, since we now basically getting the full URL and remove the query afterwards.

Be graceful with the 'baseurlpath' configuration option. We should not fail when the trailing slash is missing, just add it.

Fix another issue with URLs not being validated before displaying them into a page. Reported by John Page (hyp3rlinx).

Fix an issue with some PHP environments (mostly related to FastCGI or php-fpm, common with nginx) where $_SERVER['SCRIPT_NAME'] is already populated with $_SERVER['PATH_INFO'] appended to it. In those cases, we should not blindly append PATH_INFO to SCRIPT_NAME, but check the latter first. This hopefully resolves #5 and closes #391.

Use LONGTEXT instead of the TEXT data type in MySQL to avoid size constraints in the latter. This resolves #399.

If we have an exception registered in the state array while authenticating, we should throw that exception instead of just printing it as plain text and exit. This resolves #401.

Allow regular expression matching of trusted.url.domains.

Fixes #379

Verify that the URL passed as a parameter in the no cookie error page is allowed. Reported by John Page (hyp3rlinx).

Fix an issue with the SimpleSAML\Logger class that was still trying to fetch the track ID from the session when running from the command line, where there's obviously no session at all.

General improvements of MemcacheMonitor module

Sometimes appear in error logs as:

PHP Notice:  Undefined index: ReturnTo in .../modules/core/www/as_login.php on line 9
PHP Notice:  Undefined index: AuthId in .../modules/core/www/as_login.php on line 13

Added missing definitions

Fix missing type attribute to comply with W3c XHTML 1.0 Transitional

In the LDAP class, the password should only be escaped if it's not null, so that we don't try to bind with an empty password if none was provided. This fixes #366 and closes #370.

The data store can be a custom class in a module. When we cannot load a class matching the 'store.type' configuration option we don't handle the exception, causing a fatal error. Use the new CriticalConfigurationError to fix that. This fixes #369.

Start using the new configuration exceptions, handling error situations that before led to blank pages or even worse.

Add a couple new exceptions, both to signal configuration issues. One can be used to signal an error in a certain configuration, leaving to the user deciding whether it was critical or not, and what to do to deal with it. The other one indicates a configuration exception that we cannot recover from (i.e. missing files), and bootstraps a minimal configuration skeleton that we can use to go on momentarily.

If $_SERVER does not contain the variables we need (i.e. when running tests) it makes no sense at all to try to guess the root URI, since there's no such a thing, probably.