Tests were referring to internal implementation
details of edx-django-utils. This comment removes
those references to free the library up to be
refactored.
ARCHBOM-1584
The old folder name is somewhat confusing, because the
folder contains shims to _compensate for the removal
of sys.path hacks_, but does not contain the sys.path
hacks themselves.
Furthermore, this import_shims/ system could also be
used for other import path changes, such as turning
the locally-installed packages in common/lib/
into regular, importable modules
(e.g. `from common.lib.xmodule import abc` instead of
`from xmodule import abc`). So, a name that is not
specific to the sys.path hacks may be better
in the medium-to-long term.
Along the same lines, we also rename SysPathHackWarning
to DeprecatedEdxPlatformImportWarning.
There wasn't any reason for this to be in edx-platform in particular.
See https://github.com/edx/jenkins-job-dsl-internal/pull/444 -- we can
run `make requirements; pip install -e .` in a code-annotations
virtualenv and point it at each IDA in turn to generate reports for all of them.
ARCHBOM-1583
When module_name is None, the call to edx-django-utils'
get_code_owner_from_module crashes. So we avoid making that call when
the module_name is None, which sometimes happens (for good reasons or
not, but it's valid behaviour).
edx-toggles==2.0.0 is likely to suppress namespace objects for the
management of toggle objefcts. We explicitely prevent this by adding a
requirement constraint.
I grew tired of modifying the Makefile to compile requirements without
upgrading them. Also, installing pip-tools should not be part of the
compile-requirements target, so a separate target was created.
Part of the notifier service deprecation (DEPR-106).
Also removed pdfminer from the package uninstall list, since we no longer install the package it conflicts with either.
Passing in the INSTALLED_APPS lambda seems to result in some tasks for certain apps
like instructor_task and bulk_email to not get discovered properly.
When an InstructorTask is stuck in QUEUING (say if there was a
problem with celery), the support team needs to manually intervene
and mark the task as "FAILED" so that new tasks of that type can
be created for that course. This is usually done one at a time,
but sometimes a bug or outage might cause many tasks to fail at
once, making recovery extremely cumbersome. This commit adds the
ability to do this process in batches.
