The `_query_autostart` method in `gui.py` had inverted logic for checking the autostart status on Linux. It would return `False` when the application's path was found in the autostart file, and `True` otherwise. This caused the autostart checkbox in the settings to display an incorrect status and for the feature to not work as intended.
This commit corrects the logic by removing the `not` operator from the check, ensuring that the method returns `True` when the path is found in the autostart file, and `False` otherwise.
Now the target CPU architecture should be set by the 'ARCH' environment
variable before invoking appimage-builder.
This makes it easier to build the AppImage for other CPU architectures.
This environment variable was necessary back then to work around a bug
in the libgtk3 version we include in the AppImage.
However, since commit 3897a8bccf
("ci/linux: build AppImage on Ubuntu 22.04 (Python 3.10)"), that's not
needed anymore, because Ubuntu 22.04 has a newer version of libgtk3,
which fixes the problem.
I made sure to verify the app runs as expected on Wayland sessions of
two major desktop environments:
- GNOME: 42.9 (Ubuntu 22.04.05) and 48.3 (Arch Linux)
- KDE Plasma: 5.27.5 (Debian 12) and 6.4.3 (Arch Linux)
An up to date version of libXft is still needed because Ubuntu 22.04
still has 2.3.4, which can cause the app not to launch if an emoji font
is installed on the host.
Unfortunately, sourcing the Python venv Bash script to activate the
virtual environment doesn't work while we're building the AppImage, so
this just manually invokes Pip from within the venv to install updated
versions of meson and ninja (for building PyGObject next), and then we
prepend the venv's "bin" directory to the 'PATH' environment variable
to make meson and ninja discoverable on the next "pip install" step.
This also removes "certifi" from being manually installed, because
"truststore" has been usable since commit
3897a8bccf ("ci/linux: build AppImage on
Ubuntu 22.04 (Python 3.10)").