Fixes a crash caused by accessing a deleted view in weston_wm_window_schedule_repaint. It can be easily reproduced by switching between menus in Firefox.
Signed-off-by: Dima Ryazanov <dima@gmail.com>
The commit "Remove the weston_view.geometry.width/height fields" changed
the type of the surface configure callback function, but did not change
the callbacks in data-device.c. This commit fixes the type of the
functions left needed to be changed.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
The logind API was designed to allow any kind of devices and any number of
devices. It has no idea of "main DRM device" or similar. However, the
weston DRM backend was designed with a single DRM device as master.
Therefore, we wake it up unconditionally on session-wakeup. But this may
fail with logind as a session may be awake, but not all devices have been
resumed, yet.
Therefore, we change the weston-logind backend to deal with this case
correctly. Instead of waking up the compositor on session-wakeup, we wait
for the main DRM device to wake up. Once we get the event, we notify the
compositor.
For sleep, we reverse this logic. On *any* of the following events we
tell the compositor to go to sleep:
- Session gets inactive
- DRM device gets inactive
- DRM device is removed
This guarantees, that weston is only active if *both*, the session and the
main DRM device are awake/active.
Note that we could actually rely solely on the DRM-device Pause/Resume
events from logind and drop all the Active-Prop-Changed handling. logind
guarantees proper ordering of both. However, in case we ever change weston
to support multiple GPUs, we need the per-device notification. Thus, keep
the code. This also makes weston more fail-safe in case logind fails to
send the PauseDevice event (for whatever reason..).
The parent update on set_maximized and set_fullscreen is a behavior of
wl_shell.
That does not happen on xdg-shell, so it can't be in the set_fullscreen
and set_maximized common code, but rather in the wl_shell_surface
interfaces.
These surface types don't exist anymore inside weston desktop shell
implementation. They are just exposed as wl_shell surface types, but
internally the implementation is done with surface states.
The previous behavior (setting a surface type unsets another one) still
happens when using wl_shell. This change is mainly done as a refactory
to allow xdg-shell to use the same code.
xdg_shell is a protocol aimed to substitute wl_shell in the long term,
but will not be part of the wayland core protocol. It starts as a
non-stable API, aimed to be used as a development place at first, and
once features are defined as required by several desktop shells, we can
finally make it stable.
It provides mainly two new interfaces: xdg_surface and xdg_popup.
The xdg_surface interface implements a desktop-style window, that can be
moved, resized, maximized, etc. It provides a request for creating
child/parent relationship, called xdg_surface.set_transient_for.
The xdg_popup interface implements a desktop-style popup/menu. A
xdg_popup is always transient for another surface, and also has implicit
grab.
This has a couple of additional implications for the internal weston API:
1) weston_view_configure no longer exists. Use weston_view_set_position
instead.
2) The weston_surface.configure callback no longer takes a width and
height. If you need these, surface.width/height are set before
configure is called. If you need to know when the width/height
changes, you must track that yourself.
It is quite possible for os_create_anonymous_file() to fail when trying
to allocate a new wl_shm buffer. Propagate this failure out from
shm_surface_prepare. Most parts of toytoolkit are already avoiding NULL
cairo surfaces.
If cairo surface allocation fails, do not try to call the widget redraw
functions, those are not prepared to deal with NULL. Also do not
schedule a frame callback, this allows us to retry drawing the next
time.
If redraw fails for the main_surface of a window, restore the widget
geometry to what the compositor currently is showing. This keeps the
window visual appearance in sync with application state, so interacting
with the application does not break too badly.
If the very first draw of any window fails, then forcefully exit the
program. E.g. if weston-desktop-shell fails to allocate buffers for the
unlock dialog, w-d-s exits, and weston unlocks the screen automatically.
This patch allows e.g. weston-terminal to stop from enlarging while
resizing, if new sized buffers can no longer the allocated. Even then,
the application stays usable, as it can often repaint in the last
successful size. It does not crash, and the user is able to resize it
smaller, too.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
If posix_fallocate is available, use it instead of ftruncate. Unlike
ftruncate, when posix_fallocate succeeds, it guarantees that you cannot
run out of disk space, when later writing to the mmap()'ed file.
With posix_fallocate, if os_create_anonymous_file() succeeds, the
program cannot get a SIGBUS later from accessing this file via mmap. If
there is insufficient disk space, the function fails and errno is set to
ENOSPC.
This is useful on systems, that limit the available buffer space by
having XDG_RUNTIME_DIR on a small tmpfs.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The default can be set by passing WESTON_SHELL_CLIENT as an argument
to configure, similarly to WESTON_NATIVE_BACKEND.
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
Reset pointer button count in case the driver did not emit appropriate
number of number button released events.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
The purpose of this patch is to move away from separating evdev
functionality from udev functionality in order to rely on a separate
library to do this for us.
This patch is only compile tested and I have no idea how much it breaks.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Always put them as the top-most layer in the layer list of their parent.
This ensures that, for example, the popup menu produced by
right-clicking on a surface (which is not currently at the top of the
stacking order in the current workspace) is displayed at the top of the
stacking order.
This ensures transient surfaces are included in the layer of their
parent, even if the parent later changes layers. It achieves this by
recursively changing the layers of all children of a surface when that
surface’s layer is changed. The recursion is unbounded unless transient
surfaces are restricted to not being children of other popup or transient
surfaces.
This fixes a bug whereby a surface which is transient for a fullscreen
surface could end up being stacked below it.
https://bugs.freedesktop.org/show_bug.cgi?id=69443
It’s tied too deeply into the shell’s window stacking and ordering code
to legitimately be split out into compositor.c. Inline it in the shell,
and refactor some code around it a little, tidying up the stacking
behaviour for fullscreen surfaces.
This will be used more extensively in the next few commits, where shsurf
layering is handled more explicitly when changing the type of a surface.
This commit introduces the minor functional change that map() will now
always add the new surface to a layer list, as
shell_surface_calculate_layer_link() always returns a non-NULL link
element. This affects fullscreen and ‘none’ surfaces (which are now added
to the fullscreen and current workspace’s layer list, respectively).
This is in preparation for unifying how surface layering works. It
introduces the small functional change that fullscreen, maximized and
top-level surfaces now explicitly have no parent surface. Only popup and
transient surfaces have a non-NULL parent.
This fixes two GCC warnings when compiling with -Wswitch-enum
-Wswitch-default, and makes it clearer that those cases have been
thought about explicitly when writing the code, rather than just being
forgotten.
This fixes a load of GCC warnings when compiling with -Wswitch-enum
-Wswitch-default, and makes it clearer that those cases have been
thought about explicitly when writing the code, rather than just being
forgotten.
full screen black_surface doesn't have associated wl_buffer, so
black_surface->width and height can't get value through
weston_surface_commit(). then weston_surface_damage(black_surface)
will be wrong in shell_stack_fullscreen(), the result is that we
can't see a full screen view whe APP window's size isn't equal to
output's size like running weston-gears
Signed-off-by: Xiong Zhang <xiong.y.zhang@intel.com>
This is needed for XWayland surfaces with alpha channel, as X will be
sending crap in there that should be discarded.
This is currently done with a copy in the compositor, while we wait for
support in the VideoCore side.
Replace the two functions getting the intended surface dimensions from
the surface's buffer and buffer transformation parameters by a single
function that just set the surface size according to all the buffer
state.
The old functions were always called in pairs, and always assigned to
the surface dimension variables.
This function also deals with a NULL buffer by setting the dimensions to
zero, just like the callers used to do.
The new function has no users outside this source file, so do not export
it. This basically unexports the old functions.
This removes the calls to weston_surface_buffer_width/height() from
input.c, which are the last external calls to them.
Instead, use the cached values from weston_surface::width,height. These
have already been set by weston_surface_commit(), because that is the
only way a weston_surface can get a buffer.
Gather the variables affecting the coordinate transformations between
buffer and local coordinates into a new struct weston_buffer_viewport.
This will be more useful later, when the crop & scale extension is
implemented.
The field is already set - correctly - in the backend switch_mode.
setting output->current_mode to mode in compositor.c leads to bugs,
since mode can be freed by the shell.
For example, the shell allocates it on the stack for
WL_SHELL_SURFACE_FULLSCREEN_METHOD_DRIVER
Signed-off-by: Axel Davy <axel.davy@ens.fr>
We no longer receive an exposay_binding() call while exposay
is in-flight as we have grabbed the keyboard, so we need to
listen on the modifiers callback for the modifier press and
release.
We don't want bindings to be run while the keyboard is grabbed.
Otherwise the binding handler may grab the keyboard too, making
the old grab go away without even being cancelled.
The symbol is needed only for the EGL buffer path. If --disable-egl is
given to ./configure, there is no need for it, so fix it to actually not
look for that symbol needlessly.
This should fix the runtime error:
Failed to load module: .../rpi-backend.so: undefined symbol:
vc_dispmanx_set_wl_buffer_in_use
when you use --disable-egl and do not have a recent enough
libraspberrypi package (/opt/vc, a.k.a userland.git) that would provide
vc_dispmanx_set_wl_buffer_in_use. Apparently no released version of
userland yet provides this.
The calls are organized into two helper functions to avoid a boolean
argument, and put the #ifdefs away from the main parts of the code.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Cc: Tomeu Vizoso <tomeu@tomeuvizoso.net>
Dima Ryazanov
Jonas Ådahl
David Herrmann
Rafael Antognolli
Jason Ekstrand
Pekka Paalanen
Emilio Pozuelo Monfort
Philip Withnall
Xiong Zhang
Tomeu Vizoso
Axel Davy
MoD
Ander Conselvan de Oliveira