- Fix unclosed OBJ in core.cc when target is OpenBSD
- `to_real_path` changes:
- Canonicalize result to mimic `realpath` behavior
- Partially implement `to_real_path` for OpenBSD
- Add more tests for `to_real_path`.
- Build system improvements:
- Use same i18n include/link logic for all platforms
- Replace `HAVE_SOME_SOUNDCARD_H` and `HAVE_LINUX_SOUNDCARD_H` with `HAVE_SOUNDCARD_H`
and use other defines (__linux__) to guard include
- Replace `<XXX>_LIBRARIES` with `<XXX>_LINK_LIBRARIES` for linking.
- Libraries weren't in default linker search path on OpenBSD
- OpenBSD specific:
- Fix bsdapm.cc
- Remove diskio stats stubs
- Stub get_battery_short_status
- Remove `-stdlib=libc++` build flag on OpenBSD
Signed-off-by: Tin Švagelj <tin.svagelj@live.com>
Co-authored-by: gmb <gmb@openbsd76.my.domain>
Co-authored-by: Tin Švagelj <tin.svagelj@live.com>
We version the docs install path, but there's no reason to include the
-pre and git sha in the version for the path. With the git sha in
particular, it could result in having a bunch of dirs for each separate
git commit (annoying).
This resolves #1928.
* It was difficult/too costly ensuring temporary this was correct when vecs contained in rects were accessed directly.
* Added semantics to `rect` and some utility functions.
Signed-off-by: Tin Švagelj <tin.svagelj@live.com>
You can now specify `BUILD_COLOUR_NAME_MAP=OFF` to disable building the
gperf-based colour map, which also removes the requirement for gperf at
build time.
Rather than allowing the build to continue without gperf, we should
fail with an error so that the colour behaviour does not change in a
backward incompatible manner. The old colour behaviour should continue
to work going forward.
This resolves #1868.
- Separated displays so they don't have to all be included in build even when the features are disabled.
- Removed weird linker magic for display init and replaced it with straightforward (and faster) template functions that behave the same.
- Replaced `disabled_display_output` classes with log messages.
- Add explicit compiler errors when feature dependent headers get included where they shouldn't (wl, x11).
- Switch BUILD_MOUSE_EVENTS dependency from OWN_WINDOW to BUILD_X11.
- Other minor improvements to some existing code which shouldn't affect behavior.
- Improve documentation.
- Remove X11 from unrelated headers.
- This reaches 0% unneeded X11 polution in the sources.
- Reenabled parts of window_type which would work in Wayland.
- Remove unneeded guards.
- Display files are now no longer compiled if their features are disabled.
Signed-off-by: Tin Švagelj <tin.svagelj@live.com>
- Separate cairo-xlib.h Lua bindings from cairo.h
- This allows cairo.h to be used from Lua on Wayland without X11 as requirement.
- Add pango dependency for Wayland (adds pangocairo).
- Make graphics related bindings depend on BUILD_GUI.
- Remove BUILD_GUI guard around options in platform checks.
- Keeping only checks that **directly contain module requirements**
A way of creating surfaces in Lua that uses Shm allocated buffers should
be added in future. This commit only cleans up build options to avoid weird
errors.
Signed-off-by: Tin Švagelj <tin.svagelj@live.com>
For reproducible builds, the build date messes things up, and some
distros just set it to 1980 now anyway. Thus, it's better to remove the
date altogether.
Fix most issues with X11 event propagation.
Remove X11 code from mouse-events.cc/.h completely.
CMake:
Update toluapp cmake_minimum_required to stop it complaining during build.
Fix ConkyPlatformCheck.cmake not providing access to cairo, imlib2 and rsvg to Wayland.
Move BUILD_MOUSE_EVENTS includes under BUILD_GUI so they're not duplicated.
Signed-off-by: Tin <tin.svagelj@live.com>
