The `done` event does not emit a `closeEvent`, but saving is only
implemented in that event handler. Changing this to `close` correctly
triggers the default QWidget event and thus settings are saved.
Use the render path that preview uses for non-preview rendering, which
uses the GPU to rescale rather than the CPU and pulls the image from the
GPU before pixel format conversion happens. This means it's no longer
necessary to select BGRA to use the keyer anymore.
Mapping a GPU resource for CPU read immediately after GPU write creates
a large stall on the graphics thread. Use a small circular buffer, and
read the oldest frame instead.
The script and scene switcher lists were not being styled.
This also sets the spacing to 1 for the filter, script, scene
switcher and properties view lists, the same as other lists.
If QT_VERSION is not defined, it gets define with the AUTO value. And
its definition is moved to the helper file.
find_qt now:
- Check quietly for Qt5 and Qt6
- If QT_VERSION is set to AUTO. It checks firstly if Qt5 was found it
will use it. If not it do the same for Qt6
- If QT_VERSION is set to 5 or 6, it checks if the choice was found and
use it. And if not, it falls back to the other if found.
- If neither Qt5 or Qt6 are found, a fatal error is emitted.
- The macro saved the _QT_VERSION in the cache to replace QT_VERSION,
so the process is not repeated each time that find_qt is used.
- When Qt::Gui is in the Linux component list, Qt::GuiPrivate is added.
So using the versioned one is no longer required.
Removing the first call to `video_output_stop` prevents the `data_mutex`
field in `struct video_output` from being destroyed while still in use.
The `render_preview_source` function will call `video_output_lock_frame`
upon that mutex and encounter a NULL pointer exception.
Co-authored-by: Paul Hindt <paulh@aja.com>
Due to how CMake and generated project files are structured, just using
AUTOUIC to pick up Qt `.ui` files will lead to a situation where
changing such a file doesn't trigger a regeneration of the associated
header files and thus a re-build of the target.
Upstream fix still requires `.ui` files to be added as target sources.
CMake issue: https://gitlab.kitware.com/cmake/cmake/-/issues/17959
Removing the first call to `video_output_stop` prevents the `data_mutex`
field in `struct video_output` from being destroyed while still in use.
The `render_preview_source` function will call `video_output_lock_frame`
upon that mutex and encounter a NULL pointer exception.