It's no longer a singleton
It's now a unique_ptr that dies together with the Application
* Add getChatLogger to EmptyApplication
* unrelated change: Access Application::instance statically
* fix logging init order
* Add changelog entry
* fix: link to the current user in timeouts
* fix: `readability-suspicious-call-argument`
* docs: add changelog entry
* fix: link both users when replacing timeout
* Update changelog entry
Co-authored-by: Rasmus Karlsson <rasmus.karlsson@pajlada.com>
* Add working reconnect recent messages
* Rename method to messagesUpdated
* Use audo declarations
* Add docs to new LimitedQueue methods
* Add more documentation, try atomic loading flag
* Update CHANGELOG.md
* Remove unused include
* Rename 'reconnected' signal to 'connected'
* Reserve before filtering on arbitrary update
* Extract recent messages fetching to own class
* Use std::atomic_flag instead of std::atomic_bool
* Add PostToThread include
* Add chatterino.recentmessages logging
* Remove unneeded parameters, lambda move capture
* Remove TwitchChannel::buildRecentMessages
* Add documentation, use more clear method name
* Reword changelog entry
I think it sounds better like this :)
* Rework how filling in missing messages is handled
This should hopefully prevent issues with filtered channels with old messages
that no longer exist in the underlying channel
* Check existing messages when looking for reply
* Clean up string distribution in file
* Try to improve documentation
* Use std::function for RecentMessagesApi
* Only trigger filledInMessages if we inserted
* Remove old unused lines
* Use make_shared<MessageLayout> instead of new MessageLayout
* Alphabetize QLogging categories
* Reorder CHANGELOG.md
* Use circular buffer for LimitedQueue
* Reduce copying of snapshot
* Small optimizations
* Remove unneeded lock statements
* Add LimitedQueue tests
* Fix includes for limited queue benchmark
* Update CHANGELOG.md
* Use correct boost version iterators
* Use a shared_mutex to clarify reads and writes
* Update `find`/`rfind` to return the result as a boost::optional
* Use `[[nodiscard]]` where applicable
* Update comments
* Add a couple more doc comments
* Replace size with get
get is a safe (locked & checked) version of at
* Use std::vector in LimitedQueueSnapshot
* Update LimitedQueue benchmarks
* Add mutex guard to buffer accessors
We do not know whether T is an atomic type or not
so we can't safely say that we can copy the value
at a certain address of the buffer.
See https://stackoverflow.com/a/2252478
* Update doc comments, add first/last getters
* Make limit_ const
* Omit `else` if the if-case always returns
* Title case category comments
* Remove `at`
* Fix `get` comment
* Privatize/comment/lock property accessors
- `limit` is now private
- `space` is now private
- `full` has been removed
- `empty` now locks
* Remove `front` function
* Remove `back` method
* Add comment to `first`
* Add comment to `last`
Co-authored-by: Rasmus Karlsson <rasmus.karlsson@pajlada.com>
Adds custom channel filters complete with their own mini-language. Filters can be created in settings, and applied by clicking the three dots to open the Split menu and selecting "Set filters".
There's a Refresh button added to the popup to refresh the users messages in the popup. Not automatic now while we figure out how fast/slow it would be.
Co-authored-by: dnsge <sagedanielr@gmail.com>
"Stack" is the default behaviour, it will search 20 messages up and 5
seconds back in time to stack the timeout.
"Stack sparingly" will try to do the same, but only if the user has not
typed a message inbetween the this and the last timeout.
Fixes#1157
