Prior to this patch, the code was fetching all completion data for all
activities in a course, even when the activity was not requested. This
leads to recursion issues as the data has not been added to the cache
before this operation occurs.
To handle this situation, only the requested CM is fetched in full, and
a boolean flag is used to store whether the full data has been fetched.
When returning a partially fetched value from the cache, the flag is
used to determine whether more data must be fetched, and the cache
updated.
The flag is filtered out before the value is returned.
Note: Many of the tests were updated as these were inspecting private
features of the API which should not really be tested.
Activity modules may not have the associated grade_item created yet. It
used to throw fatal error in that case - even when trying to view the
course or edit the activity. So there was no easy way to recover from
this situation.
The patch is based on reasoning that an activity without grade item is
same as activity without any grades. And as such it is considered
incomplete.
A new unit test is added to cover this specific scenario. The existing
unit test is modified and it does not expect the exception any more.
There does not seem to be any good reason why this situation should be
exceptional.
When an automatic completion condition may have had its state change,
we now unset the cached value for the user's completion in the relevant
activity, so up-to-date values are re-fetched and available to students.
The previous behaviour was that custom conditions would remain cached
until the activity reached overall completion.
We should be able to fetch the grade completion status for modules
that do not have custom completion configured. This also improves
unit testing coverage.
* Fixed inverted $sameuser test data.
* Fixed caching expectation check. Caching only relies on whether the
user accessing the completion data is the same user or not.
* Fixed checking for the caching of other modules. Should have been
checking cm ID and not instance ID.
* Additional test case when whole course parameter is passed as
true, but the requesting user is different from the target user.
I've gone over a few of the mofified files (those
which were showing warnings and errors to CiBoT. Some of them
have been fixed completely, while others only have fixed
for the lines belonging to this issue (lib/tests/moodlelib_test.php)
for example.
The current ->setMethods() has been silently (won't emit any
warning) in PHPUnit 9. And will stop working (current plans)
in PHPUnit 10.
Basically the now deprecated method has been split into:
- onlyMethods(): To point to existing methods in the mocked artifact.
- addMethods(): To point to non existing (yet) methods in the mocked
artifact.
In practice that means that all our current setMethods() calls can be
converted to onlyMethods() (existing) and done. The addMethods() is
mostly useful on development phases, not final testing.
Finally note that <null> isn't accepted anymore as parameter to
double all the methods. Instead empty array [] must be used.
Link: https://github.com/sebastianbergmann/phpunit/issues/3770
Mocke at() matcher is being deprecated with phpunit9 and
will be removed with phpunit10.
Source: https://github.com/sebastianbergmann/phpunit/issues/4297
Luckily we are using those deprecated matchers only in completionlib
tests, so there aren't many cases to modify. Now, we are using
supported matchers (once, exactly, never...) and the tests have
been reorganised to better represent the expected behavior (how
many times stuff is called, with which params and return values).
Use the custom completion implementation for mod_choice to test
completion_info::get_state() to cover the case where the completion
state is being determined from the custom completion condition.
Move the current logic for determining the completion status for the
"Student must receive grade" completion rule to a function so it cann
be reused.
Unit test included.
* Unit tests for completion_info::get_data() and
completion_info::internal_get_state are mocked which causes failures
with the new implementation. It's more straightforward and realistic
to generate real course and modules to test these methods.
Both assertContains() and assertNotContains() are deprecated in PHPUnit 8
for operations on strings. Also the optional case parameter is. All uses
must be changed to one of:
- assertStringContainsString()
- assertStringContainsStringIgnoringCase()
- assertStringNotContainsString()
- assertStringNotContainsStringIgnoringCase()
More info: https://github.com/sebastianbergmann/phpunit/issues/3422
Regexp to find all uses:
ag 'assert(Not)?Contains\('
Changes:
- Activities with auto completion and a completion status overridden to
COMPLETION_COMPLETE are no longer processed by normal completion
triggers.
- All activities can still be completed by students when their
completion status has been overridden to COMPLETION_INCOMPLETE, via
either auto or manual triggers.
- Completion unit tests updated
The completion cache is currently not marked as simpledata. On the
course page it is frequently retrieved hundreds of times which results
in many calls to the slow unserialize function. By making a slight
change to the data format (using arrays instead of objects) we can
mark it as simpledata, which will avoid using unserialize.
1. getMock()
2. setExpectedException()
3. checkForUnintentionallyCoveredCode renamed to beStrictAboutCoversAnnotation
4. beStrictAboutTestSize renamed to enforceTimeLimit
5. UnitTestCase class is now fully removed.
instance is has been used instead of cmid. This works as all id's are 1
in most cases, but not always.
accesslib_clear_all_caches_for_unit_testing was updated to clear the
$USER->access cache which would fail test when the wrong user's data was used.
