made changes to test with previous version of django-waffle
updated the query count to test
testing with version 0.13
testing with version 0.14
testing with version 0.15
added version 0.14
updated the django-waffle version to use 0.18
updated the django-waffle version to use 0.16
updated the query counts to pass tests
ran make upgrade
updated the django-waffle to support django 2.2
made changes to test with previous version of django-waffle
updated the query count to test
testing with version 0.13
testing with version 0.14
testing with version 0.15
added version 0.14
updated the django-waffle version to use 0.18
updated the django-waffle version to use 0.16
updated the query counts to pass tests
removed the pdb statements
ran make upgrade
This adds request caching to the following places:
* course expiration wrapper (displayed in Units)
* offer banner generation (displayed in Units)
* get_enrollment
* user_by_anonymous_id
* youtube_disabled_for_course
On a sample course with edx-val enabled, this reduced the queries
for a large sequence from 450 to 155.
Specifying a namespace in django.conf.urls.include() without providing an app_name is deprecated.
Adding the app_name attribute in the included module.
The MockS3Mixin prevents the correct setup of the ModuleStoreTestCase
and made this test fail. Since the fix for this wasn't trivial, this
test was skipped on python 3, and now is removed.
A user's last logged in value previously wasn't updated when the user
logs in through the oauth2 flow from the ios mobile app. Here we must
send the user_logged_in signal manually. It's implemented in the
my_user_info mobile rest api endpoint because adding it to where the
oauth2 flow happens is to complex, and the mobile app hits this endpoint
after a successful login anyway.
https://github.com/boto/boto/issues/2868
In python 3 contentfile can act like a file but return unicode strings.
Boto doesn't like that and it causes issues when encoding the bytes for
transfer to s3.
When we run the management command, the photo_id_key is retrieved
through the orm and so is represented as a string. However, when we do
the initial attempt, the object is instantiated and we refer to the key
without having decoded it. The photo_id_key is sent as a part of the
request to software secure. And if we have a byte representation their
signature checking will fail.
This is why the management command worked even when the site didn't.
It was getting a cleaned key via the orm.
There were cases where we needed to encode things to codecs other than
ascii. In these cases, python returns byte strings but we needed them
to be unicode so that they serialize correctly later when we combine
them with other unicode strings.
