The behat hooks were needlessly complicated which made them much harder
to read, and understand, leading to bugs during development.
These have been significantly simplified to favour clarity over
overloading.
The session was previously only started once, but has now been moved to
restart between each test. Because of this the `$session->start()` is
now called more frequently which exposes a pre-existing issue whereby
Selenium will periodically fail to start.
When the session was only started at the beginning of the test, and at
occasional other times, it made sense to always stop the test runner
because it signified an issue with obtaining the session at all.
Since the session is restarted more frequently, and now between every
test, it no longer makes sense to do this. The test should only be
_stopped_ if the first attempt to obtain the Session fails. Once a
Session has been retrieved at least once it no longer makes sense to
assume that the failure was a configuration failure, but a connection
failure. Therefore subsequent tests should be attempted.
Some browsers like Firefox are very inflexible with window.open()
and block it if it is not instantly invoked after the user click.
Also according to https://stackoverflow.com/a/6807615 it is best
practice to replace self:: with static::
This feature adds an admin tool for creating custom licenses.
Now custom licenses can be added and amended in Moodle, and the site
default can be set to a custom license.
Core licenses remain hard-coded and are uneditable, so they will always
require update within Moodle core updates, and maintain their
internationalisation through core language strings.
This also includes fundamental changes to the license API including
the addition of license caching and deprecation of no longer required
admin settings for license management.
In the get_grade_letters there is a static variable that is used
to cache grade letters, we should use MUC for this so that it is
reset properly between unit tests.
Previously, if the RSS feed (extests) was being served locally,
CFG->proxybypass was applied and causing the test to fail (proxy
ignored).
In that exact test we need the configured proxy to be applied
always, because the test is exaclty about detecting the problem
when a wrong proxy is used.
