When building from a release tarball, there is no .git directory
and git rev-parse would either fail or pick up a parent repository's
HEAD. Guard the git version detection with fs.is_dir('.git') so
that tarball builds produce a clean version string.
Requires importing the fs module.
Signed-off-by: Rudi Heitbaum <rudi@heitbaum.com>
(cherry picked from commit e5db9506b4bfff602b704329d5b7ff0a8a6bdb42)
These handlers aren't invoked for all types of xdg_shell surfaces:
they are only invoked for xdg_toplevel, and are not for xdg_popup.
Rename them accordingly.
If no application is specified, run without starting a child
process. Leave it up to the user to start applications externally
and stop cage with a signal.
For bug reports, it's useful to ask for a debug log. However there's
no way for users to enable debug logs without recompiling. Add a
CLI flag to do so.
Instead of waiting for the surface to be mapped before sending a
configure event, do it on initial commit.
Note, wlroots 0.18 no longer sends automatic configure events on
initial commit.
This is unnecessary because Xwayland logic doesn't bring in new
dependencies. We can just compile in our Xwayland-related code
when wlroots has been built with Xwayland support.
See this Sway patch: https://github.com/swaywm/sway/pull/8165
wlr_scene_output_layout_add_output() aborts when called with an
already-added output.
To reproduce, run wlr-randr to reconfigure one of the enabled
outputs.
When destroying an xwayland surface, the dissociate and destroy handlers
are called, but both of these were removing the map and unmap signal
handlers, causing a segfault when the destroy handler went to remove
them.
Fixes#309
wlroots as a subproject now yields the xwayland meson option to its
parent. We need to match the type for this to work. This also adds
support for auto mode, where xwayland is used if present but no warning
is given otherwise.
The output is disabled then re-enabled using the wlr-output-management
protocol with tool such as wlr-randr.
$ wlr-randr --output HDMI-A-1 --off
$ wlr-randr --output HDMI-A-1 --on
When re-enabled the new output configuration is committed before the
output to be added to the global output layout. However,
handle_output_commit() expects the output to be already part of this
global layout and assert failed. Now, do not treat this in
handle_output_commit() and simply ensure the resulting scene output is
not NULL in handle_output_frame().
Otherwise, testing to set preferred / "best" mode will always return an
error. Consequently, output will simply be configured with last and
probably "worst" mode.
Only terminate if the last output was nested under the Wayland or X11
backend. If not, using DRM backend for example, terminating Cage when
unplugging the last monitor or simply turning it off does not seem to be
the right behavior.
wl_touch::frame() is expected to be sent to client to indicate end of
touch frame event and not sending it may cause issues.
For example, Qt applications using Qt Wayland platform plugin do not
consider touch events until this end of frame to be received.
Some applications indicate different shutdown conditions by returning
specific exit codes. One of these is e.g. Kodi, which returns 64 in case
the user chose "Power off" and 66 in case the user chose "Reboot".
In order to act on these exit codes, it thus makes sense in some
situations to pass them on from the primary client to the caller of
Cage.
This exit code is only returned if the primary client is the cause of
Cage terminating.
Co-authored-by: Patrick Steinhardt <ps@pks.im>
Previously, the output of `meson build` would warn about the use of
some functions that have been deprecated since meson v0.56.0, and
the use of the `check` kwarg in the `run_command()` function.
Update meson.build to fix these build warnings.
- Update wlr_box includes to util/box.h: the wlroots header has
been moved upstream.
- Subsurface fields have been moved
- Create renderer and allocator, stop using wlr_backend_get_renderer
- Initalize output rendering
This fixes the drag icon not disappearing when releasing the
pointer button.
Reproduction steps:
- Open gedit
- Drag some text around
- Release the pointer button
view.c:238:52: error: no member named 'subsurfaces' in 'struct wlr_surface'
wl_list_for_each (subsurface, &view->wlr_surface->subsurfaces, parent_link) {
~~~~~~~~~~~~~~~~~ ^
/usr/include/wayland-util.h:443:30: note: expanded from macro 'wl_list_for_each'
for (pos = wl_container_of((head)->next, pos, member); \
^~~~
/usr/include/wayland-util.h:409:32: note: expanded from macro 'wl_container_of'
(__typeof__(sample))((char *)(ptr) - \
^~~
Based on https://github.com/swaywm/sway/commit/3162766eef14
cage.c:307:59: error: too few arguments to function call, expected 2, have 1
server.backend = wlr_backend_autocreate(server.wl_display);
~~~~~~~~~~~~~~~~~~~~~~ ^
subprojects/wlroots/include/wlr/backend.h:43:21: note: 'wlr_backend_autocreate' declared here
struct wlr_backend *wlr_backend_autocreate(struct wl_display *display,
^
output.c:456:36: error: no member named 'transform' in 'struct wlr_output::(anonymous at /usr/local/include/wlr/types/wlr_output.h:154:2)'
wl_signal_add(&wlr_output->events.transform, &output->transform);
~~~~~~~~~~~~~~~~~~ ^
Based on https://github.com/swaywm/sway/commit/2c76923282b1
xdg_shell.c:230:2: warning: implicit declaration of function 'wlr_xdg_surface_for_each_popup' is invalid in C99 [-Wimplicit-function-declaration]
wlr_xdg_surface_for_each_popup(xdg_shell_view->xdg_surface, iterator, data);
^
ld: error: undefined symbol: wlr_xdg_surface_for_each_popup
>>> referenced by xdg_shell.c:230 (../xdg_shell.c:230)
>>> cage.p/xdg_shell.c.o:(for_each_popup)
Based on https://github.com/swaywm/sway/commit/5438cc158a1b
cage.c:307:61: error: too many arguments to function call, expected single argument 'display', have 2 arguments
server.backend = wlr_backend_autocreate(server.wl_display, NULL);
~~~~~~~~~~~~~~~~~~~~~~ ^~~~
/usr/include/sys/_null.h:34:14: note: expanded from macro 'NULL'
#define NULL ((void *)0)
^~~~~~~~~~~
/usr/local/include/wlr/backend.h:36:21: note: 'wlr_backend_autocreate' declared here
struct wlr_backend *wlr_backend_autocreate(struct wl_display *display);
^
Based on https://github.com/swaywm/sway/commit/53f5197c26f1
seat.c:263:56: error: use of undeclared identifier 'WLR_KEY_PRESSED'
if ((modifiers & WLR_MODIFIER_ALT) && event->state == WLR_KEY_PRESSED) {
^
Based on https://github.com/swaywm/sway/commit/bb342ac5e6de
This restores the original behaviour from 2cf40f7, is easier to read,
and satisfies clangformat.
This also doesn't call wlr_idle_notify_activity anymore, which was
probably unintended.
../output.c:235:57: error: no member named 'base' in 'struct wlr_buffer'
wlr_output_attach_buffer(wlr_output, &surface->buffer->base);
~~~~~~~~~~~~~~~ ^
In this mode, only the last connected output will be used. If that one
is unplugged, the previously last connected output will be enabled. This
for example allows one to switch between two outputs, e.g. on a handheld
device such as a mobile phone.
This enum provides two means of behaviour for multi-output setups:
extend the display across all outputs, or only use the last one. The
former is the current (and default) behaviour; the latter will be added
in the next commit.
These will allow different patterns of multimonitor behaviour. In a
followup commit, we will introduce a behaviour where only the last
connected output is in use.
* Don’t terminate display when no view is found
Some apps like RetroArch will quit and fork itself to switch
modes (like from the menu to a loaded core). This means that for a
very short period of time we have no view available for Wayland. Right
now, Cage doesn’t actually exit when it does this terminate, so you
get this kind of zombie child process that is running but not showing
anything on the screen because there is no compositor.
The solution I have here is to just keep Cage’s Wayland server running
to avoid this issue. I’m open to other idea, but this seems sane to
me. Perhaps an alternative is to check if the process is still alive
when this happens, if not then we can do the quitting behavior. In
addition, we could make this an option if some users don’t want this
behavior.
* Remove "ever_been_mapped" from cg_wayland_view
If something goes wrong during startup (i.e., the wlroots backend cannot
be created), we jump to the end. This then unconditionally removes the
sigchld_source, but that hasn't been set at this point yet.
* Implement support for wlr_keyboard_group
* Cast pointers to void type when passing to wlr_log
* Fix bracket style
* Improve style
* Remove unnecessary function
* Prefer defined variable for brevity
* Remove unnecessary pointer printing
* Ameliorate style
* Remove duplicate functionality
* Set group repeat info
* Ameliorate style
* Fix bug introduced through renaming
* Ameliorate style in seat.h
* Remove cg_keyboard in favour of cg_keyboard_group
* Remove unused signalling
* Apply clang-format
* Fix ordering of commands
* Remove unnecessary field in cg_keyboard_group
This adds a first iteration of a clang-format style definition. It
formats the current code such that it aligns (mostly..) with the style
I prefer, but it is most likely incomplete and there will be corner
cases.
Fixes#124.
Previously, is git was found but the build was run from a snapshot,
`version` wouldn't be set. This commit fixes this by first setting
version to the project version and only then, conditionally, setting it
to the git version.
Fixes#125.
With the new CLOEXEC setting, spawning the primary client
can fail *after* forking. In this case, the client process
has been forked and will need to be cleaned up.
In case something fails before pid has been set, it's set to 0.
From waitpid(2):
The value of pid can be:
0 meaning wait for any child process whose process group ID
is equal to that of the calling process at the time of the
call to waitpid().
That will be none in this case, and hence this won't block and is thus
safe.
As mentioned by @emersion:
By default, pipe creates FDs without the CLOEXEC flag set, which means
they will be leaked to any other child process spawned. Would be nice to
set the CLOEXEC flag to prevent the leak.
wl_display will destroy the outputs, whose destroy handler will remove
them from the output layout. But by that point, the output layout has
already been destroyed.
According to the Wayland docs:
If the event source is registered for re-check with
wl_event_source_check(): 0 for all done, 1 for needing a re-check. If
not registered, the return value is ignored and should be zero.
See e.g. http://manpages.ubuntu.com/manpages/cosmic/man3/wl_event_source.3.html
Since we don't register any of these for re-checking, we should
return 0.
This prevents a segfault when shutting down during startup, when a
client is passed that doesn't spawn a window.
It also brings us one step closer to not having to have a pointer to the
backend in cg_server.
Fixes#70. For development, initialise meson with the `--werror` option:
```
meson build --werror
```
Alternatively, for an existing build, run
```
cd build
meson configure -Dwerror=true
```
With Cage now supporting hotplugging of outputs, we shouldn't warp the
cursor to the center of every new output. Rather, we should warp it only
on the initial startup.
If the dialog extends the output layout in at least one dimension, then
the part of it that is larger than the output will be rendered
of-screen. This commit maximizes such dialogs instead.
Fixes#110.
The function also did an intersection check, so it did more than just
getting the surface box. This refactoring makes it more clear.
This commit also fixes a bug in coordinate spaces that makes dialogs
render properly over multiple outputs.
This is based on code from Sway, which is also MIT licensed hence
compatible. This makes the surface damaging and rendering code easier to
follow and makes it easier to import future changes to Sway as well.
Following breaking changes in wlroots 0.9.0, wlr_output_commit must be
called after wlr_output_set_mode, wlr_output_set_transform and wlr_output_enable.
Additionally, wlr_output_enable is no longer implicitly called by
wlr_output_set_mode.
Fixes#102
Outputs are arranged in a horizontal layout in the order they are
created in by wlroots. Maximized xdg_shell views will span all outputs,
like the global fullscreen mode in sway.
Fixes#87
A new tag for packagers to build Cage with wlroots 0.6.0.
Changelog:
* Update to wlroots 0.6.0
* Don't submit too much damage in damage tracking implementation
* Use XKB_* environment variables to configure keyboard devices
On outputs that have modes, we need to set one before we can use it. We
pick the preferred mode, or if this is not advertised, the last listed
mode. This mode is generally the mode with the highest resolution.
Fixes#10
Since we now have a tag with 0.1 and also a release with that version
number was published on GitHub we should adjust the version in the meson
file too.
GTK still uses the (now deprecated) server-decoration protocol, which is
the predecessor to xdg-decoration. Hence, with this commit Cage now also
hides decorations on applications such as Firefox.
Fixes#47, see also
https://github.com/Hjdskes/cage/pull/45#issuecomment-466402865.
This commit adds a commandline switch (-d) to disable client side
decorations, if possible. In this case, Cage will not draw any
decorations of its own, in order to maximize screen real estate.
The default behavior remains the same, i.e., if -d is not passed,
clients will draw their client side decorations, if any.
Fixes#32
Cage uses wl_event_loop_add_signal to handle SIGINT and SIGTERM, which
masks these signals. This means that the subprocess spawned by Cage
start with these signals masked, which can lead to delays in Cage
shutting down on e.g. ^C. Hence, we now unmask all signals between fork
and exec.
Fixes#40
Override-redirect windows are meant to be ignored by the X server, at
least as far as position goes. Hence, we detect whether we're dealing
with such a window and consequently don't position it.
This makes, amongst others, Chromium's popup menus and dmenu appear on
the correct position and size.
Otherwise, we'll segfault:
0x00007ffff7c36715 in __strlen_avx2 () at /usr/lib/libc.so.6
0x000055555555ae00 in view_get_title (view=0x5555558403c0) at ../view.c:34
This was temporarily removed when introducing proper view abstractions.
Now that we have the cg_xwayland_view struct, we reintroduce this
workaround in its proper place.
This also fixes the previous workaround, which checked whether a view
was *currently* mapped and not if it *has ever* been mapped.
See #18, #19 and 443d955dfd.
When using the Wayland or X11 backend, Cage is drawn inside a window.
This commit sets this window's title to that of the currently focused
toplevel window inside Cage.
Fixes#29.
This is to work around "misbehaving" (for lack of a better term)
clients. At the moment, Firefox Nightly and Google Chrome/Chromium are
known to be suchs client:
When XWayland support is enabled, Firefox first creates an XWayland
surface, closes this (before mapping it) and then opens an XDG toplevel
surface. Cage tries to manage the first XWayland surface, but when it
closes, Cage has no surfaces left and hence closes as well.
Hence, Cage terminates before it picks up on Firefox's XDG toplevel, and
Firefox (rightly) prints it cannot read the Wayland pipe.
In Chromium's case, it simply opens an XWayland surface which it
immediately closes, before opening the "real" XWayland surface.
The workaround is to track whether an XWayland surface has been mapped
and, if it hasn't, to not exit when we have no views left.
Firefox's behavior and the workaround are discussed in #18.
This commit fixes#18 and is part of the fix for #19.
This is the path we settled on in #24.
That is: any new toplevel window takes over the Cage display, hiding any
previous toplevels until it is closed. Only when the last toplevel is
closed, does Cage exit as well.
This allows us to check per-view whether is has dialogs open, instead of
diong it on a global basis as we are doing currently. This is necessary
for fully supporting multiple primary clients.
With this, some apps (e.g. mpv) can block the idle tracker from
kicking in. This way, the screen won't blank (or whatever you
configured) when the application doesn't want it to.
With Cage becoming more popular since its mention on Phoronix and
therefore getting more use-cases than just my own project, add XWayland
support. The refactoring of 2cf40f7 makes this much easier. Note that
this is a no-cost addition for those of us not using XWayland as it is a
compile-time option that needs to be explicitly enabled by adding
`-Dxwayland=true` to your meson command.
We shouldn't render a window before it is mapped (obviously), but we
render all windows in the view list. Hence, only insert the window once
it is mapped.
We could run into the case where a window is destroyed without being in
the window list, so we now track unmapping again and remove windows from
the list when they get unmapped.
Since this is inherently output independent, we can move this to here
and avoid the inexistance of an output we ran into in the previous
commit.
Warping the cursor is no problem here either: since we restrict
ourselves to a single output, there won't be any confusing UX by having
the cursor jump from one output to the newly attached one.
This makes Cage much easier to maintain. Not only is it easier where to
look and to maintain a mental model of the code, there is also more
encapsulation, better abstractions and better extendability.
Since we do not ever minimize or otherwise unmap clients (except when
destroying them), we don't need to track mapped/unmapped state. The
destroy case is captured in checking whether the surface has a buffer
attached to it, before requesting its texture.
This makes it easier to extend in the future. Also, it will make it
easier to add (un)install targets and eventually, distribute files such
as systemd services.
It appears that wlr_xdg_toplevel_set_maximized does not actually change
the size of the surface. Rather, it only makes the surface think it is
maximized. Hence, we need to set the size to that of the output
manually.