| Age | Commit message (Collapse) | Author |
|---|
|
Since the old keys used to be strings i started out by
json serializing the new keys with type, value, rev.
However storing serialized json in couch (json) does
not really make sense. So now we do not serialize but
instead have one json document. The lookup for a key of
type pgp may still return a string but for everything
that uses the new api it will return a hash with type
and revision.
This data structure is way easier to handle also on the
nickserver side.
|
|
|
|
|
|
We used to just ignore the key.
Also separated the code for handling key updates from other
user updates. This should eventually be moved to a different
route. Mixing the two makes the implementation really hard.
|
|
This simplifies returning the user while still working
on it a lot. Much cleaner than all these return user statements.
There's a lot more to refactor here. For example delegating methods
to user etc. ... but for now this should suffice. Don't want to
break this in a bugfix release.
|
|
fixes #8807
|
|
They used to be sorted by the code which was not helpful
fixes #8806
requires deploy of new design docs to the platform
|
|
Production instances are getting cluttered with invites from
test accounts. Instead of marking them as used we will now completely
remove them.
refers to #8804
refers to #8807\
|
|
|
|
make the rescue in account creation less broad
See merge request !15
|
|
Related with https://github.com/pixelated/pixelated-user-agent/issues/924
With @aarni
|
|
Related with https://github.com/pixelated/pixelated-user-agent/issues/924
With @aarni
|
|
It was catching errors that should have led to 500s
|
|
If a login was invalid as a username but also for the identity we used to have duplicate error messages. Let's avoid that.
Also added a test to make sure invite_code errors are properly displayed no matter what other fields are set or missing. Pixelated will rely on this to test invite codes
|
|
We moved it into the lib folder so it's not in an autoload path anymore.
So now it needs to be required before being used.
This fixes a load order issue that would cause non-deterministic failures
in CI
|
|
use_database affects all uses of prepare_database - so also the one
in tmp_database. In order to avoid that we do not use_database but
just overwrite the database method itself.
|
|
We already did the same for other concerns. The way we load models
for couchrest migrations does not work well with concerns in the
model directory as they will be loaded twice.
|
|
for him/herself
So that it we do not expose the is_admin property to anyone else
including other admins.
|
|
feature/expose_admin_in_api
|
|
|
|
We saw errors from duplicate loading of LocalEmail and
LoginFormatValidation. The latter resulted in a crash.
In an attempt to ensure all subclasses of Couchrest::Model::Base are
loaded Couchrest::Model::Utils::Migrate requires all files in app/models.
We have an extension that does the same for the engines.
During this process LoginFormatValidation and LocalEmail were autoloaded
when 'identity' was required. Afterwards they were required again.
It looks like rails' autoload mechanism does not play nicely with require.
So to make sure they are not autoloaded first move the concerns and
helper classes into the lib directory and require them explicitly.
|
|
So that whoever consumes the API can use this attribute to
determine if admin functionalities should be made available to
the current user.
|
|
This expresses the intent rather than the implementation.
Also replace temp with query refactoring.
|
|
|
|
CouchRest > 1.2 does not use RestClient anymore. So we should not try to
catch its errors.
|
|
|
|
create/delete test & tmp users.
|
|
configured auth token.
|
|
token, "tmp" users are users that exist only in tmp db, "test" users are either tmp users or users named "test_user_x"
|
|
are configured in the static config, to be used for infrastructure monitoring.
|
|
the only race condition I can think of here is this...
somebody tries to authenticate with a token that is almost expired.
auth checks and notices it is not expired yet so starts to prolonge it.
Before the polonged token is written to the db the cleanup script discovers that it has just expired.
prolonged token is written to the db
cleanup script fails to delete it from the db as it has been updated.
So what we want in this case is to keep the token alive as it was renewed in the last minute.
|
|
get re-enabled.
|
|
anymore. closes #7690
|
|
|
|
https://github.com/Alster-Hamburgers/leap_web into remove_couponcode
# Conflicts:
# lib/tasks/invite_code.rake
|
|
Braintree_implementation
|
|
Id is set = invite code to make it easier to delete the invite codes that get created during the leap platform tests (because this avoids having to look up ttheir id each time).
|
|
Replacing the Coupon Code gem with base64/securerandom as suggested by Elijah.
|
|
Id is set = invite code to make it easier to delete the invite codes that get created during the leap platform tests (because this avoids having to look up ttheir id each time).
|
|
Replacing the Coupon Code gem with base64/securerandom as suggested by Elijah.
|
|
|
|
This will simplify invite code deletion after running tests.
|
|
Removing some superfluous code, mostly, and structuring tests better.
|
|
|
|
|
|
|
|
|
|
The rake task now takes a second (optional) argument that sets the number of uses per invite code.
If this is omitted, the default number of uses is 1.
(This commit also contains some minor code cleanup that removes some stuff that I'd commented out but not removed.)
|
|
Introduce a invite_max_uses property to invite codes to allow admins to set a maximum number of uses for invite codes.
|
|
Through the config param 'invite_required', providers can decide whether users need to provide an invite code upon signup.
The default setting is false.
|
