This replaces a :disabled selector with individual ones for each
element type; I verified in CSS selector profiling that this change
removes lines that were previously about 10% of profiled CSS selector
processing time for rendering the combined feed.
The reason being that the rendered elements have very few disable-able
elements, but the browser was spending a lot of time trying to match
**every** element for whether it might have been disabled, which was
slow.
This probably reflects a browser bug.
Adds note about feature being available only to organization owners
and administrators.
Fixes the icon in the note about BigBlueButton to be the zulip icon
that's currently used in the compose box.
This commit updates the stale logic in 'toggle_topic_visibility_policy'
function. It was not updated after the addition of 'follow' visibility
policy resulting in buggy behavior.
Earlier, 'Shift + M' hotkey was working in unsubscribed channels.
This commit updates the logic to make it non-functional. This
helps to keep the UI consistent as we don't offer a way to
change visibility_policy in unsubscribed channels via
recipient_bar_controls or other UI elements.
Note: We don't implement this restriction in server because
we don't delete the user-topic states in db when a channel is
unsubscribed as it helps to retain those states when re-subscribed.
And implementing this restriction at server level can create
a confusion of why such states exists in db when API doesn't allow.
Even though the border-color was set for `#subscriptions_overlay
.subsection-parent div`, since the border-style was not set. It did not
have any effect on the appearance of the children div of
.subsection-parent.
These were added in ae7fe85ec5, and
should have been deleted in 333b8b095c.
Having those rules did not make any difference to the subscription and
user group list items.
Having the `div` selector there affects the performance. See
https://chat.zulip.org/#narrow/stream/6-frontend/topic/CSS.20selector.20performance/near/1845719
for more details.
We've introduced a new class called `.client-logos-div`.
In `hello.ts` we just manually append the name of this new class for the
logo changing logic since that felt more readable than adding the class
programmatically.
What was happening:
* Focus being in the search input text box made Up/Down be handled
by the browser and move the cursor. This is default browser behavior,
and not useful to us (you can use home/end for the same goal, but
also one just doesn't put much text in that input given the pills
system).
* We probably don't want to push keyboard focus out of the search box
after completing a search, since that would make it annoying to type
a new search term or whatnot.
The fix is to just make the existing Up/Down keyboard handler apply
even if the search input is focused.
More context here:
https://chat.zulip.org/#narrow/stream/9-issues/topic/Can't.20navigate.20search.20results.20using.20the.20keyboard.20.2331291/near/1913535
The class was named `typeahead-option-label` for the element containing
the actual option label. We renamed it to
`typeahead-option-label-container` to give it a more appropriate name.
Zopfli[^1] performs very good, but time-intensive, zlib compression.
It is hence only suitable for pre-compressing objects, not on-the-fly
compression.
Use a webpack plugin to write pre-compressed versions of JS and CSS
assets using Zopfli, and configure nginx to serve those assets when
`Accept-Encoding: gzip` is provided.
This reduces the size of the JS and CSS assets on initial pageload
from 1422872 bytes to 1108267 bytes, or about a 22% savings.
[^1]: https://github.com/google/zopfli
This parameter is unused for Zulip avatars, and results in a smaller
(pixel-wise) Gravatar than we use for Zulip "small" avatars. Omitting
the `s=50` parameter results in a 80x80 image, which is closer to the
100x100 image that Zulip thumbnails avatars to. Ironically, using the
default (larger) Gravatar dimensions also results in _smaller_
responses (in bytes) from Gravatar.
We are using `.val` to set compose box content which is very fast vs
`setFieldText` which is very slow due to it doing a lot of
forced repaints. The major downside of using `val` here is that
user will not able to perform `undo` operation on this which doesn't
seem something user would want to do here.
Note that this effects compose content restored from drafts,
scheduled messages and on reload.
Similar to zulip.mp3 in the previous commit, this file is corrupted
and only playable in some browsers.
Re-download it from the source[^1], and convert it to MP3 via
`ffmpeg`.
[^1]: https://freesound.org/people/InspectorJ/sounds/411089/ per docs/THIRDPARTY
This file's bytes changed in ba46dc83c6, corrupted by some unknown
process; its bytes begin:
```
00000000: efbf bdef bfbd efbf bd64 0000 0000 0000 .........d......
00000010: 0000 0000 0000 0000 0000 0000 0000 0000 ................
00000020: 0000 0000 0000 0000 0000 496e 666f 0000 ..........Info..
00000030: 000f 0000 000f 0000 1a1e 0011 1111 1111 ................
00000040: 1122 2222 2222 2222 3333 3333 3333 4444 ."""""""333333DD
00000050: 4444 4444 4455 5555 5555 5555 6666 6666 DDDDDUUUUUUUffff
00000060: 6666 7777 7777 7777 77ef bfbd efbf bdef ffwwwwwww.......
00000070: bfbd efbf bdef bfbd efbf bdef bfbd efbf ................
00000080: bdef bfbd efbf bdef bfbd efbf bdef bfbd ................
00000090: efbf bdef bfbd efbf bdef bfbd efbf bdef ................
```
This is not an MP3 file (which should begin with the bytes `fffb`) nor
an MP3 encapsulated in ID3 (which would begin `4944 33`). This left
it playable by some browsers, but not others like Safari.
Restore the version from before ba46dc83c6, which is universally
recognized as an MP3 file.
Earlier, we were using 'send_event' in
'notify_realm_custom_profile_fields' which can lead to a situation,
if any db operation is added after the 'send_event' in the action
functions using it, where we enqueue event but the action function
fails at a later stage.
Events should not be sent until we know we're not rolling back.
Fixes part of #30489.
The database operations in 'access_user_group_for_setting' and
'check_add_user_group' used in 'add_user_group' view should be
collectively atomic.
This commit adds transaction.atomic decorator for that purpose.
Earlier, we were using 'send_event' in 'delete_user_grou' codepath
which can lead to a situation, if any db operation is added after
the 'send_event' in future, where we enqueue events but the
function fails at a later stage.
Events should not be sent until we know we're not rolling back.
Fixes part of #30489.
Earlier, we were using 'send_event' in 'edit_user_group' codepath
which can lead to a situation where we enqueue events but the
function fails at a later stage.
Events should not be sent until we know we're not rolling back.
Fixes part of #30489.
Tim Abbott
Lauryn Menard
Prakhar Pratyush
Alex Vandiver
Vector73
Aman Agrawal
Shubham Padia
evykassirer
Varun Singh
Karl Stolley