Prior to this, the search is still racy but one tends to notice
this only when the DB is large or network is involved.
The user can initiate a search, get bored, navigate to another chan
issue a different search.
Now however, the results of the first search come back in and
hilarity ensues as we are now confused with the state.
To avoid this, keep track of the last search done and any result
that comes in that isn't equal to the active query is garbage and
can be dropped.
The only thing that cares about user colors is the user component.
Putting a class value on the chat component seems to be the wrong
place.
This also allows us to remove various css selectors so that we
don't need to be that specific.
After all whatever has that class needs to be colored, we don't
care where it is.
Offset is eventually passed to sqlite as an OFFSET clause.
This works as follows:
sqlite> select num from seq limit 5 offset 0;
┌─────┐
│ num │
├─────┤
│ 1 │
│ 2 │
│ 3 │
│ 4 │
│ 5 │
└─────┘
sqlite> select num from seq limit 5 offset 5;
┌─────┐
│ num │
├─────┤
│ 6 │
│ 7 │
│ 8 │
│ 9 │
│ 10 │
└─────┘
However, the code currently emits a request for offset + 1, which ends
up skipping a message
sqlite> select num from seq limit 5 offset 5+1;
┌─────┐
│ num │
├─────┤
│ 7 │
│ 8 │
│ 9 │
│ 10 │
│ 11 │
└─────┘
Nachtalb put some infra in place that was never actually working.
It errors out when a user clicks on a message.
Remove the offending code, but keep it all in place so that we
can improve on it.
When we hit doSearch, we always reset the offset value to 0,
meaning we always hit the conditional (!0) and always set the
messageSearchInProgress flag to undefined.
This is wrong, we do want to set this flag when we initiate a search.
Firefox does not seem to select leading (or trailing) characters that are
too small; so this commit sets a very small width, that is still large
enough to be selected.
This commit also adds `display: inline-block`, so the width is not
ignored; but this causes Chrome to ignore the space after `>`, so I made
it a non-breakable space.
An alternative is to make only the leading `only-copy` an
`inline-block`, but I think the non-breakable space is a good idea
regardless.
Update the code in MessageCondensed that generates the condensed
messages ("X users have joined, Y modes were set") to count the number
of actual mode changes instead of the raw count of MODE messages. One
mode message can contain multiple mode changes.
Signed-off-by: Taavi Väänänen <hi@taavi.wtf>
window.event is a deprecated global that's set to the currently
dispatched event.
- Opened and closed mentions box by clicking its icon in the top bar
- Left and right clicked on an inline channel name and saw context menu
open both times
- Two-finger swiped on iOS and saw channel change
- Long-touched and dragged channel in network list on iOS and reordered
the list successfully
This commit adds a new command, /kickban, that is a combination of /kick
and /ban: it kicks the specific user from the channel and then sets the
+b mode to ban the user from the channel.
* Extend test coverage to the `search` function.
* Test sort order of messages from `getMessages` and `search`
* Move reversal of `search` results from Vue to messageStorage.
* Remove unnecessary uses of `sqlite.serialize` in tests.
* Return promises from test functions where possible.
Chrome seems to somewhat often auto fill the text input of the
highlight exception list with my username as the next field that
follows is of type password.
Try to work around that by telling chrome not to autofill either of
those.
Do note that this is only a hint... The broser vendors apply some
$magic heuristics and if they trigger they ignore the hint.
Every time the component was mounted it would add another listener.
Since old listeners would often error this could cause a lot of log
spam, particularly when using the hotloader on a mobile device.
