If a user scrolls just below the 1000px threshold and clicks the back-to-top button within 250ms, we'll never hide the button as the scroll event gets throttled away. That's rather inconvenient.
closes#7729
When there were posts with many followed tags they were returned
multiple times, resulting in less than 15 unique posts. That resulted in
some posts to be missed in the stream.
Fixes#4503closes#7715
We only return `nil` when we know the person, but don't know the key or
the key is invalid, so it doesn't make sense to retry in this case. When
the person isn't known and can't be fetched we raise a DiscoveryError
which will be retried.
Also the errors were moved to the `Signable` module in the last release.
closes#7717
This is to let the author of the post know, that this pod is interested
in updates about this post.
The sending user is only used to verify that the participation was sent
from this pod, but lets use an admin/podmin account if available.
closes#7708
Some people may remove their birthday date after the notification was
sent, which then breaks the notification page for other users. Let's
just display the date when the notification was created, and not display
the users updated birthday date. When users update from date A to B it
always looks weird anyway, when we display the same new date B twice on
different days, or display two different dates for the same user.
We could remove notifications when users change or remove their
birthday, but that would be way more complex and also we usually don't
remove notifications (not even for deleted posts).
Fixes#7689closes#7691
- Removed posts and non contacts from other's data
- Collections are exported in batches to lower memory footprint
- In base exporters create User object instead of keeping instance because it caches all associations
closes#7627
It is not included into the main test suite, because it has :performance tag.
One can run this test using command:
$ bin/rspec --tag performance spec/integration/export/memory_usage_spec.rb
This test creates additional fixtures set to speed up repeated runs.
Main purpose of fixtures for us is to speed up test runs. Rebuilding
fixtures each run makes them pointless. They will be rebuilt automatically
if some of "files_to_check" are changed, so we're protected from the case
when fixtures are outdated and don't load properly.
It was possible to run migration locally without providing old private
key. This way migration was performed but not dispatched, which obviously
leads to desynchronization of the federation state so let's validate sender
before performing any actual actions.
closes#7558
The locals option is only supported in ActionView::TestCase, so it is
ignored anyway (even worse, it disabled the complete assertion). I added
some more assertions to ensure the template is rendered with the correct
conversation.
closes#7588
This was in all stack traces, which confused people ...
We currently only support IE11, so the version check was already old
again ...
When we break old browser versions, we don't check that for other
browsers. So older IE browser just break, like very other older browser
too.
closes#7557
This commit introduces support for AccountMigration federation message
receive. It covers the cases when the new home pod for a user is remote
respective to the recepient pod of the message. It also allows to initiate
migration locally by a podmin from the rails console. This will give the
pods a possibility to understand the account migration event on the
federation level and thus future version which will implement migration
will be backward compatible with the pods starting from this commit.
This doesn't add any value in the post preview:
* Avatar is already at the left-hand side of the publisher
* Timestamp is always "less than a minute ago"
* Interactions are disabled in the preview anyway
We don't display these things in the comment preview either, they only
need space and don't add any value to the preview.
