In the near term, we wish to precisely preserve the existing values of
all Django settings exposed by lms/envs/production.py in order to avoid
breaking legacy Django plugins without a proper announcement.
That includes preserving the behavior of these old, redundant dicts:
* ENV_TOKENS
* AUTH_TOKENS
* ENV_FEATURES
* ENV_CELERY_QUEUES
* ALTERNATE_QUEUE_ENVS
Particularly, it means we need to ensure that updates to Django
settings are reflected in these dicts. The most reliable way to do that
is to change the yaml-loading logic so that these values are aliased
to the corresponding values in the global namespace rather than
deep-copied.
Finally, we remove KEYS_WITH_MERGED_VALUES from the global namespace,
and inline the remaining list. We have modified the list (specifically,
we dropped the no-op MKTG_URL_OVERRIDES). Plugins should not be counting
on the value of the list, so we remove it.
This reintroduces commit 15939232d5,
which was reverted due to a typo.
The typo is fixed in the commit immediately following this one.
Co-Authored-By: Feanil Patel <feanil@axim.org>
The `dump_settings` command currently prints out the raw `repr(...)`s for
defined functions, e.g.:
"WIKI_CAN_ASSIGN": "<function CAN_ASSIGN at 0x74ce5e9b2020>",
In addition to being uninformative, these `at 0x74ce...` hashes change every
run, so they appear in the diff as having "changed" every time. With this
commit, here's what `dump_settings` will print out for a function instead:
"WIKI_CAN_ASSIGN": {
"module": "lms.djangoapps.course_wiki.settings",
"qualname": "CAN_ASSIGN"
},
`notify_credentials` has 2 ways of running.
1. The manual, one-off method which uses `--args_from_database` to specify what should be sent.
2. [The automated method](7316111b35/openedx/core/djangoapps/credentials/management/commands/notify_credentials.py (L157-L159)), which runs on a regular schedule, to catch anything which fell through the cracks.
The automated method does a certain amount of time/date math in order to calculate the entry point of the window based on the current runtime. This is, I assume, why it has some hardcoded logic; it's not at all simple to have a `cron`-run management command running on a regular cadence that can do the same time logic.
```py
if options['auto']:
options['end_date'] = datetime.now().replace(minute=0, second=0, microsecond=0)
options['start_date'] = options['end_date'] - timedelta(hours=4)
```
However, it is not ideal that the actual time window of 4 hours is hardcoded directly into `edx-platform`.
This fix
* creates an overridable `NOTIFY_CREDENTIALS_FREQUENCY` that defaults to 14400 seconds (4 hours).
* pulls that frequency into the quoted code
Using seconds allows maximum flexibility.
FIXES: APER-3383
This command dumps the current Django settings to JSON for
debugging/diagnostics. The output of this command is for *humans*... it
is NOT suitable for consumption by production systems.
In particular, we are introducing this command as part of a series of
refactorings to the Django settings files lms/envs/* and cms/envs/*.
We want to ensure that these refactorings do not introduce any
unexpected breaking changes, so the dump_settings command will both help
us manually verify our refactorings and help operators verify that our
refactorings behave expectedly when using their custom python/yaml
settings files.
Related to: https://github.com/openedx/edx-platform/pull/36131
* fix: catch a possible exception in beta course configuration
when the learner is a beta tester, and the beta test has been set up with a large duration, the courseware djangoapp can return an overflow error on a timedelta call. In those circumstances, this defaults to returning an unmodified date.
FIXES: APER-3848
Adds the concept of "downstream-only" fields to the XBlock upstream sync logic.
Downstream-only fields are customizable fields set only on the downstream XBlock -- we don't keep track of the upstream field value anywhere on the downstream XBlock. Changes made to these fields in the upstream block are ignored, and if the link to the upstream block is severed, the downstream changes are preserved (not reset back to defaults, like the upstream-tracked customizable fields are).
The fields chosen as "downstream-only" are those related to scoring and grading.
The `max_attempts` field was previously a customizable field that tracked the upstream value. However, because it is scoring-related, it has been converted to a "downstream-only" field.
This change impacts course authors' use of library content in their courses.
* feat!: Removing the long-deprecated legacy course_modes chooser
`course_modes/choose.html` (and its corresponding
`_upgrade_button.html`) were specifically only used for the edge case
where an enterprise user found themselves in the non-enterprise learner
dashboard, and attempted to enroll in a course outside of the
enterprise flow. Once upon a time, in a 2U-only workflow, the commerce
system would apply specific discounts for users within the said case.
That's no longer true, and it has never been true outside of this one
company.
Removing this template cleans up a legacy version of a legacy page that
was, realistically, exclusively seen by employees of 2U,
and nobody else.
Removes:
* The corresponding testsfor behavior only seen in the legacy page.
* A waffle flag since all cases route as if the flag is set: `VALUE_PROP_TRACK_SELECTION_FLAG`: `course_modes.use_new_track_selection`
* Some variables set in `CourseModeView` which were only ever rendered in the legacy template (`title_content`, `has_credit_upsell`) have been removed from the class.
* There is a high likelihood that the class is still a target for re-factoring now that the legacy view is gone, but I'm hesitant to touch something which is not covered by previously existing tests, because the logic around what template gets rendered when is complex.
FIXES: APER-3779
FIXES: #36090
* build: Drop an unused Dockerfile
We never cleaned up the docker file we were using for custom github
runners when we switched to standard runners. This cleans that up and
removes another Docker mention that is out of date.
Co-authored-by: Kyle McCormick <kyle@axim.org>
Returning the list with a 'completion' string was added in 9bc0f85. However,
the `get_accessible_sequences` method expects a list of dicts with an 'id' key.
This should improve performance for courseware operations when there is
content that doesn't map to any existing XBlocks in the system. This
usually happens to old courses when custom XBlock types are deprecated
and removed, or when a course is imported from another instance with a
different set of installed XBlocks.
We would like new devstack instances to better support the user retirement pipeline out of the box after provisioning. The first step is to update the default list of retirement states devstack is configured to add/use.
fix: enterprise-customer-support should only return learners from given customer
Commit generated by workflow `openedx/edx-platform/.github/workflows/upgrade-one-python-dependency.yml@refs/heads/master`
The CourseQualityView used to call edx-val's get_video_for_course(),
which would return a fully serialized data structure that included all
encodings and inefficiently serialized them with many n+1 queries. This
is tolerable in a paginated web view, but not when pulling all of a
large courses's videos at once.
Making this change collapsed the number of queries for a large sample
MIT course from over 3000 down to 1.
This change disables the profiling panel for performance reasons. It's
rarely useful anyway, given the lack of granularity in the data it
displays.
This commit also enables the Cache panel, which is import for tracking
where we're making excessive calls to redis/memcached.
