[3.x] Added primary clipboard for Linux

[ConteZero]

This commit add support for primary clipboard on Linux to handle text paste with middle mouse button.
This is a backport of #53702 to 3.x branch.

[akien-mga]

This should also include changes from follow up PRs (at least #54045, I didn't check if there was any other), otherwise it will exhibit the same issues.

[ConteZero]

3.x branch does not have DisplayServer has_feature method so I've used some ifdef to solve #54045
I think it's the more safe and less invasive way to solve the problem.

[akien-mga]

3.x branch does not have DisplayServer has_feature method so I've used some ifdef to solve #54045 I think it's the more safe and less invasive way to solve the problem.

There is OS::has_feature which I think would be better for this than yet another define.

The command line is pretty long already :)

g++ -o platform/x11/os_x11.x11.tools.64.o -c -std=gnu++14 -Wctor-dtor-privacy -Wnon-virtual-dtor -Wnoexcept -Wplacement-new=1 -g3 -ggdb -Wno-error=return-type -pipe -fpie -Wall -Wextra -Wwrite-strings -Wno-unused-parameter -Wno-misleading-indentation -Wshadow-local -Walloc-zero -Wduplicated-branches -Wduplicated-cond -Wstringop-overflow=4 -Wlogical-op -Wattribute-alias=2 -Werror -DDEBUG_ENABLED -DDEV_ENABLED -DNO_EDITOR_SPLASH -DTOUCH_ENABLED -DALSA_ENABLED -DALSAMIDI_ENABLED -DPULSEAUDIO_ENABLED -D_REENTRANT -DJOYDEV_ENABLED -DUDEV_ENABLED -DX11_ENABLED -DUNIX_ENABLED -DOPENGL_ENABLED -DGLES_ENABLED -D_FILE_OFFSET_BITS=64 -DPTRCALL_ENABLED -DTOOLS_ENABLED -DMINIZIP_ENABLED -DZSTD_STATIC_LINKING_ONLY -DGLAD_ENABLED -DGLES_OVER_GL -DHAVE_MNTENT -Ithirdparty/freetype/include -Ithirdparty/libpng -Ithirdparty/glad -Ithirdparty/zstd -Ithirdparty/zlib -Iplatform/x11 -I. platform/x11/os_x11.cpp

Unlike DisplayServer::has_feature, it relies on strings, so I think you'd have to add logic to OS_X11::_check_internal_feature_support like this:

diff --git a/platform/x11/os_x11.cpp b/platform/x11/os_x11.cpp
index 7127447f14..4af06351a2 100644
--- a/platform/x11/os_x11.cpp
+++ b/platform/x11/os_x11.cpp
@@ -3357,7 +3357,17 @@ Error OS_X11::shell_open(String p_uri) {
 }
 
 bool OS_X11::_check_internal_feature_support(const String &p_feature) {
-	return p_feature == "pc";
+	if (p_feature == "pc") {
+		return true;
+	}
+
+#ifdef __linux__
+	if (p_feature == "primary_clipboard") {
+		return true;
+	}
+#endif
+
+	return false;
 }
 
 String OS_X11::get_config_path() const {

That's assuming that primary clipboard is a Linux-only feature and not something provided by *BSD systems too. Otherwise, you could then indeed just rely on the already existing X11_ENABLED define.

[ConteZero]

I've removed the DEFINE

[akien-mga]

I'm still not sure why it's Linux specific though, as opposed to X11 specific. The implementation for primary clipboard doesn't make a distinction between GNU/Linux and FreeBSD/OpenBSD/NetBSD.

To clarify, the platform is called x11 on 3.x and linuxbsd on master because it's not only a GNU/Linux platform port. The same code (modulo a handful of defines) can compile on other Unix + Freedesktop platforms that support X11 such as several *BSD distributions.

[ConteZero]

I'm still not sure why it's Linux specific though, as opposed to X11 specific. The implementation for primary clipboard doesn't make a distinction between GNU/Linux and FreeBSD/OpenBSD/NetBSD.

To clarify, the platform is called x11 on 3.x and linuxbsd on master because it's not only a GNU/Linux platform port. The same code (modulo a handful of defines) can compile on other Unix + Freedesktop platforms that support X11 such as several *BSD distributions.

I can only test on Linux so I really don't know if the primary clipboard code works on *BSD.
We can remove the
#ifdef __linux__
if some *BSD user can confirm that it works.

[akien-mga]

Searching "OpenBSD X11 primary clipboard" gives a number of results which seem to infer that it does work. If it doesn't work, the X11 API would return an error and hopefully we already handle it gracefully anyway.

[ConteZero]

Searching "OpenBSD X11 primary clipboard" gives a number of results which seem to infer that it does work. If it doesn't work, the X11 API would return an error and hopefully we already handle it gracefully anyway.

OK, I've removed the #ifdef

[akien-mga]

Thanks!