I don't think there's necessarily anything wrong with using non-ASCII
characters, since UTF-8 is basically the de facto standard. There's no
reason to mix a small number of curly quotes into a configuration file
though, so I've replaced them for consistency.
This fixes a lot of the issues with the LocalUser changes that were
merged recently. There was a problem where the attributes of LocalUser
were not being eagerly loaded and because the Session was detached an
exception was being raised when they were accessed.
This also fixes some typo's which were introduced.
Finally this adds a temporary fix for a potential SQLAlchemy bug, this
is a bug where doing:
User.query.filter(LocalUser.username == "some_username").first()
does NOT yeild a user with the username "some_username" but all users
on the site. The temp fix is to just query the LocalUser, this should
be resolved when bug is confirmed and fixed upstream.
This introduces a migration which adds a dummy Client, RequestToken
and AccessToken. These are used when an invalid request comes in,
instead of bailing early, it needs dummy data to prevent timing
attacks.
This then implements the methods which get the IDs of the dummy
objects. If these are changed in the future a migration which checks
for the previous dummy object should be created and updates them to
reflect the new IDs/tokens.
It seems that the GET params on a URL should be included when
one signs the request. Mediagoblin was just using the base URL
without them. This should fix that.
After the recent model changes there were some bugs which were
introduced into the serialization methods of the models. This commit
fixes those issues.
This makes the changes needed for federating MediaEntry objects as well
as adding the migration and necessary methods to get the public_id just
in time (JIT).
The code base had many references to User.username and other
specific to LocalUser attributes as that was the way it use to exist.
This updates those to query on the generic User model but filtering
by attributes on the LocalUser.
This adds the two new user models (LocalUser and RemoteUser) to the
MODELS list that is in models.py. This stops the strange bug that occurs
if you migrate a fresh database, the two models don't exist however
migrating an existing database would create them as the migrations
exist.
The code base had many references to User.username and other
specific to LocalUser attributes as that was the way it use to exist.
This updates those to query on the generic User model but filtering
by attributes on the LocalUser.
This adds the ability to search for any user based on the generic
User case and be given back the specific LocalUser or RemoteUser.
This will require any code using the model to look which attributes
they are searching on and specify the specific User model they are
on if they're not on the generic User model. This will also require
new users to be created with LocalUser.
mock recently dropped Python 2.6 support and then
re-introduced it in its latest release (however,
it's a bit unstable right now).
I think at this point, we can be more cautious
and just use mock==1.0.1.
mock recently dropped Python 2.6 support and then
re-introduced it in its latest release (however,
it's a bit unstable right now).
I think at this point, we can be more cautious
and just use mock==1.0.1.
The migration had a problem where other tables still referenced the migration
as well as a typo in an earlier migration. They have both been fixed and tested
on PostgreSQL and SQLite3.
This also fixes a bug where sometimes when creating an activity it'd raise an
Exception as the object hadn't got an ID. This has been fixed globally with a
fix to the create_activity federation tool.
