We’re getting unexplained exceptions from URL.createObjectURL in
Firefox 88 (#18374); contain them to this function.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Previously, the hook to save drafts when closing the compose box was a
focusout on #compose-textarea. This mostly worked, and was essentially
invisible to users (because the animation to notify you about saved
drafts was in the close_compose_ui widget that is hidden), but you'd
end up seeing the tooltip appear (in the wrong place) if you closed
the compose box immediately after sending a message with Tab+Enter.
The correct fix is for the drafts hook to be just before we start
clearing compose box state, inside hide_box.
This was difficult to catch in a development environment, since one
doesn't use that flow with "Enter sends" enabled.
Cleaning up test_realm_domains.RealmDomainTest.test_list_realm_domains,
test_subs.StreamAdminTest.test_private_stream_live_updates,
test_subs.StreamAdminTest.test_realm_admin_can_update_unsub_private_stream
and test_subs.StreamAdminTest.test_non_admin_cannot_access_unsub_private_stream.
We were apparently not using the ListWidget API for replacing the
content of a widget without removing the whole thing from the DOM and
replacing it.
Unless ListWidget has unexpected bugs, this should have the exact same
result as the previous logic, with much a nicer user experience.
We use styling same as that of the number shown for overflow
(e.g "+3") which has a gray background.
To keep these avatars up-to-date, we need to rerender the
recent topics view after receiving a `muted_users` event.
Also update the user documentation to mention this detail.
We deliberately avoid mentioning "recent topics" because
this applies to mobile too.
This seems more consistent with what users would expect; Recent topics
may be a better default view in general, but if a user has picked
another default view, we should use that where the default view is
desired.
This new function optimizes how we fetch subscriptions
for streams. Basically, it excludes most long-term-idle
users from the query.
With 8k users, of which all but 400 are long term idle,
this speeds up get_recipient_info from about 150ms
to 50ms.
Overall this change appears to save a factor of 2-3 in the backend
processing time for sending or editing a message in large, public
streams in chat.zulip.org (at 18K users today).
The default label for empty narrows depends on whether it's a
stream/topic narrow or a PMs narrow.
We leave some default text in compose.hbs for reply label
because it take some time for the js to display the
correct label.
We change the title to be more descriptive and also link to the
overall Zulip changelog, since one can reasonably get to this page
when googling for "Zulip changelog".
The browser seems to autofill the `Name` field of the add-playground
form. Most likely this behavior is a result of value of the `name`
input field being `name`, causing the browser do to something weird
here. This name is now changed to `playground_name`.
If the caller has already fetched the Stream or subscription details
for the user, those can be passed to has_message_access to avoid extra
database queries.
These thresholds are in relationship to the
`autovacuum_freeze_max_age`, *not* the XID wraparound, which happens
at 2^31-1. As such, it is *perfectly normal* that they hit 100%, and
then autovacuum kicks in and brings it back down. The unusual
condition is that PostgreSQL pushes past the point where an autovacuum
would be triggered -- therein lies the XID wraparound danger.
With the `autovacuum_freeze_max_age` set to 2000000000 in
`postgresql.conf`, XID wraparound happens at 107.3%. Set the warning
and error thresholds to below this, but above 100% so this does not
trigger constantly.
The old logic, inline in the compose area, has produced a very weird
effect where the buttons would move to fit the notification, ever
since design changes to use the full bottom row space.
We address this by just using a Tippy tooltip instead.
When the format of the response received from the outgoing webhook
server is invalid (unparsable json, or just wrong format that doesn't
translate into a dictionary etc.), a message with the error is sent to
the bot owner. We should include the actual payload to make reasonable
debugging possible.
In notify_bot_owner we have to move the `if response_content` block to
append the payload to the message whenever it was specified as an
argument to the function. It shouldn't be nested inside
`elif status_code` as before.
This removes a possible window where an installer error could leave
`nvm` in a state where it had prepended the full path to the
newly-installed `npm` to `$PATH`; we would like to avoid `nvm`
fiddling with path whenever possible (ref ebe930ab2c).
Share code between `safe_full_names` and `get_display_full_names`
functions, and rename `safe_full_names` to `get_full_names_for_poll`,
because that was the only place where this was used earlier.
This also has the nice side effect of showing "Muted user" instead
of the muted username in poll results.
This makes it parallel with deliver_scheduled_messages, and clarifies
that it is not used for simply sending outgoing emails (e.g. the
`email_senders` queue).
This also renames the supervisor job to match.
The reason for this flake is it missed clicking on the Set
All button (`.subs_set_all_users`) because it calculated
its position before/during it starts scrolling.
There are two scrolling events caused by typing `ot`,
* First is due to internal focus call before typing
which scrolled down the page to bring input in the view.
* Second is after it typed `ot` the user list is updated to
one value, which caused the modal height to decrease rapidly.
Note: It theory, there is three scrolling event,
First is of course internal focus then, after it typed `o`
the list is updated to two values and one is after `t` the
list is updated to a single value.
But as puppeteer is very quick it directly jumped to the
scrolling event after it typed `ot`.
For a more detailed explanation read,
https://chat.zulip.org/#narrow/stream/43-automated-testing/topic/master.20failing/near/1173996.
This commit just temporarily fixes this flake, it's not the
best approach to use time-based delays as they can't be
robust (e.g If the machine is super slow then time-based
delays can fail.).
Instead of changing the hash to the default_view hash, we render
the default_view directly in case of recent topics and all messages.
This also fixes the bug that user is unable to go back in
browser window history.
We don't want to rely on browser hash to check if RT is visible
because soon we want to display default_view without any hash
and RT may or may not be visible in that hash.
Anders Kaseorg
Alex Vandiver
Tim Abbott
Gaurav Pandey
Steve Howell
Vishnu KS
Abhijeet Prasad Bodas
Aman Agrawal
Sumanth V Rao
Mateusz Mandera
Riken Shah