Miniaudio now runs everything in a separate audio thread - this uses boost::asio's io_context.
Our miniaudio implementation is now also much simplified - it does not use its own resource manager or device. This might end up being stupid if sounds don't work after changing output device or locking or w/e
I've made the sound controller into an interface, meaning we can support multiple sound backends in Chatterino. I've added a Null sound backend that disables all sound. A QMediaPlayer or QSoundEffect or Qt backend could be added.
Miniaudio might idle & disable the device now too, not sure
I've added some unrelated changes in the form of a new setting type, and a new setting page helper function for it, which will hopefully make adding new enum settings easier in the future.
This setting stores its value as a string instead of an int, and uses magic_enum to convert between that string value and its enum value.
* Moved implementation of the methods to the `cpp` file.
* Added `DebugCount::Flag(s)` and `DebugCount::configure(name, flags)`.
* Moved from `QMap` to `std::map` (order is important here).
* Used `QStringBuilder` for concatenations.
* Used `QLocale` for formatting (adds separators).
* Added `DebugCount::Flag::DataSize` for data sizes in bytes (and fixed language to English).
* Used `DataSize` for image sizes (maybe this should be moved somewhere else?).
* Added copy button to popup.
* Fixed Image usage reporting being eight times too large (could be another PR, but honestly it's four characters).
The rendering of selections was not aligned to the actual selection that took place for newlines at the end of messages, if they were the only part that was selected of that message.
In addition to that fix, we've already refactored the MessageLayoutContainer to try to make it a little bit more sane to work with in the future.
Co-authored-by: Rasmus Karlsson <rasmus.karlsson@pajlada.com>
This code has been sitting behind a define since 2018 (de3a490). I'm not opposed to this code coming back however in this state, it feels better to remove it.
* Remove restriction on go to message on system messages
Fixes#4608
* changelog
* Change the changelog entry from a bugfix to minor
---------
Co-authored-by: Rasmus Karlsson <rasmus.karlsson@pajlada.com>
* Fix tooltip & popup positioning
This tries to ensure the tooltip & popups are created on the correct
monitor
* Add changelog entry
* Clean up debug output
* Use the full frame geometry to figure out screen bound movements
* Remove the now-unused `setStayInScreenRect` function
* Change the UserInfoPopup offset to be based on its width & height
instead
* Remove more debug output
* Fixed tab completion rarely completing the wrong word.
Fixes: #3101
* Use QSignalBlocker instead of janky bool, add comment about hidden logic
* copypasteo
* refactor: remove singletons from message rendering
* chore: add changelog entry
* Disable the `cppcoreguidelines-avoid-const-or-ref-data-members` clang-tidy check
* auto *app
* Selection is a struct, not a class
* Use ChannelView's `signalHolder_` instead of `channelConnections_`
* Remove `applySettings` step, instead just connect & set each setting individually
* rename & constify some context values
* Handle empty "last message color" setting value better (as it was
originally in this pr before I removed that change :-)
* unrelated mini refactor cleanup
* let painSelection handle size_t instead of int
* Add some more comments to the MessageLayoutContext structs
---------
Co-authored-by: Rasmus Karlsson <rasmus.karlsson@pajlada.com>
* refactor: move ipc queue into its own class
* refactor: move windows.h related functions to AW
* refactor: make NM-Client methods static
* refactor: json access
* refactor: use struct initializer
* refactor: move `handleMessage` to anon-namespace
* refactor: clean-up includes
* refactor: move action handler to functions
* refactor: cleanup `handleSelect`
* fix: cleanup clang-tidy warnings
* chore: simplify json
* revert: keep handlers as methods
This is more readable and extensible.
* fix: typo
* fix: namespace
* fix: rename define
* refactor: `IpcQueue` to be simpler
* fix: rename cmake option
* fix: use variant when constructing
* fix: make it a ref
* fix: its a pair now
This fixes a crash that could occur when closing a split before the
display name had a chance to update
The reason I found this was because the LiveController change I made
updated display names more regularly
Additionally, we now make sure to not send duplicate `displayNameUpdated` signals
upon each request for channels with CJK characters in their display name
* Default-initialize the `actualDisplayName` with the user's login name to not send an initial display name update if the display name is the same casing as the login name
* Specifically, this adds a caller to the network request, which makes the
success or failure callback not fire.
This has the unintended consequence of the block list not reloading if
the usercard is closed, but it's not a big concern.
* Add unrelated `-DUSE_ALTERNATE_LINKER` cmake option
From 0517d99b46/CMakeLists.txt (L87-L103)
Themes are loaded from the Themes directory (under the Chatterino directory, so %APPDATA%/Chatterino2/Themes).
Themes are json files (see the built in themes as an example).
After importing a theme, you must restart Chatterino for it to show up in the settings
The scripting capabilities is locked behind a cmake flag, and is not enabled by default.
Co-authored-by: nerix <nerixdev@outlook.de>
Co-authored-by: pajlada <rasmus.karlsson@pajlada.com>
The `SettingsPage::onShow` handler that sets up the block
list is only called when the application settings dialog
has already been shown once before. This caused the blocked
user list to appear empty until the second time it was opened.
Correct this by explicitly regenerating the block list in the
constructor of `IgnoresPage`.