The intended behavior is that a quick click (press and then release
within 500ms) just pops up the menu and doesn't select anything. Then
we can mouse around and and click to select an item. Alternatively, a
click and hold (ie press and release after 500ms) lets you press right
button, mouse down on the menu item you want and release to select it.
This is how menus work in most toolkits.
The handling in weston is fine, it's there to handle the case where
the button release happens outside any client window, since the client
doesn't get those events. If such a release happens late or we get a
second release outside the popup window we shut down the popup.
The problem is in toytoolkit, where we need to select the item if we
get a release within 500ms or if we get a second release. A second
release is the case where the first release came after 500ms and
didn't pop down the menu, and the second release event is from a click
on a menu item.
https://bugs.freedesktop.org/show_bug.cgi?id=52456
Grabbed widgets should always receive motion events as if it was the
widget that would receive it if no grab was active. This means that the
focused widget should always be passed as the widget argument to widget
motion handlers.
This reverts commit 8c9c8fcf6e.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
The simple clients all just call wl_display_dispatch() in a while loop
without checking the return value. Now, if the server dies or other
error occurs, we get a -1 return value instead and need to break the loop.
configure.ac: The toytoolkit clients used to get libEGL linked to them
even if there was no cairo-egl. This is useless, and actually harmful on
platforms, where libEGL absolutely requires one of the GL ES libraries
to be linked in, too.
Look for EGL-related packages only for cairo-egl with toytoolkit.
window.c: protect all GL header includes with HAVE_CAIRO_EGL, since that
is the only case we can support EGL, GL, or GLESv2 at all. In the case
we do not have cairo-egl, add enough definitions to let us build the
stubs for EGL-related functions.
Remove some #ifdefs that were inside of the same #ifdef already.
These changes allow to build sorfware rendering toytoolkit clients
without any bits of EGL libs or headers.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
If the keyboard modifier event was received after the key event the
modifier state would end up incorrect.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Callbacks registered via display_set_output_configure_handler() are
promised to be called when we know the current mode for the output. If
the following order of events happens:
1. toytoolkit binds to a wl_output global
2. application registers an output configure handler
3. the wl_output.mode events are received
Then in step 2 we would call the callback with uninitialised output
informations, giving it a 0x0 size.
To avoid such race, do not call the callback from
display_set_output_configure_handler() if the output has 0x0 size.
The wl_output.mode event will be received later, and that will trigger
the right call to the callback.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Apply wl_surface.frame request only on the next wl_surface.commit
according to the new protocol.
This makes it explicit, which repaint actually triggered the frame
callback, since commit schedules a repaint. Otherwise, something causing
a repaint before a commit could trigger the frame callback too early.
Ensure all demo clients send commit after wl_surface.frame. Note, that
GL apps rely on eglSwapBuffers() sending commit. In toytoolkit, it is
assumed that window_flush() always does a commit.
compositor-wayland assumes renderer->repaint_output does a commit.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Make input region double-buffered as specified in the new protocol.
While doing it, get rid of the undef region code, and instead use a
maximum sized real pixman region. This avoids special-casing regions
that might sometimes be undef.
As the input region is now usable by default instead of undef,
weston_surface_update_transform() does not need to reset the input
region anymore.
weston_surface_attach() no longer resets the input region on surface
size change. Therefore, also weston_seat_update_drag_surface() does not
need to reset it.
Update toytoolkit to set input region before calling wl_surface_commit()
or swapBuffers (which does commit).
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Make wl_surface.set_opaque_region double-buffered as required by the new
protocol. Also, do not reset the opaque region on surface size changes
anymore. Only explicit requests from the client will change the region
now.
In clients, make sure commit happens after setting the opaque region.
Mesa does not need a fix, as it never touches the opaque region.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Use wl_surface_commit() to commit the buffer attach, as Weston now
requires.
NOTE: GL-applications are broken until you upgrade to a version of Mesa
which does wl_surface_commit() on eglSwapBuffers(). If you have
Cairo-gl, this means all toytoolkit apps, too.
simple-shm and simple-touch OTOH will work now.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add THEME_FRAME_MAXIMIZED flag so the theming system can know not to draw
shadows for maximized windows. This allows maximized surfaces' content to be
sized and placed in a more expectable fashion.
The workspace state parameters were initialized after the first
roundtrip. If a workspace manager state event was received during this
roundtrip the state parameters were cleared leaving an incorrect state.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
desktop-shell never returned from display_run() since it
was essentially killed when weston exited. To fix this,
it is necessary to watch for EPOLLHUP in window.c so that
toytoolkit clients will return from display_run() when
weston quits. This allows for clients to clean up
as needed.
Signed-off-by: U. Artie Eoff <ullysses.a.eoff@intel.com>
Two buttons are added to the right-click menu of the window frame for
moving a surface either up or down.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
The correspondence between cursor functions and names of cursors has
never been standardized. As a consequence, each cursor function can be
represented as a cursor with one of several names. Be more robust when
loading cursor by trying all known names that correspond to a cursor.
This should fix https://bugs.freedesktop.org/show_bug.cgi?id=50487
and https://bugs.freedesktop.org/show_bug.cgi?id=52609 a bit more
thoroughly.
E.g. this can happen when you grab the lower right corner of a window
and move over the top of the window when resizing. In this case, the
changed width is still important and should be acted upon.
https://bugs.freedesktop.org/show_bug.cgi?id=53560
This patch, along with the wayland patch, adds the ability to specify
a cursor theme in the weston.ini file:
[cursors]
theme=THEME_NAME
If specified, than Weston can use a specific X cursor theme for the
pointer. This relies on the 0001-Add-support-for-X-cursor-themes.patch
for wayland.
[krh: edited to use shell section and key name cursor-theme]
We default to setting the minimum size to the initial size. To set a
different minimum size than the initial size, set the minimum size first
then then initial size. Good enough for a toy toolkit.
https://bugs.freedesktop.org/show_bug.cgi?id=50263
It is possible that a client loses the focus between receiving a
pointer.enter event and sending a pointer.set_cursor request. In that
case, the cursor surface might not be mapped and the frame callback
requested on it will never trigger.
Work around this by trying to remap the cursor surface whenever there
is a frame callback and the serial for the enter event is higher than
the cursor serial.
window.c:1173:6: warning: ignoring return value of ‘read’, declared with attribute warn_unused_result [-Wunused-result]
desktop-shell.c:305:6: warning: ignoring return value of ‘read’, declared with attribute warn_unused_result [-Wunused-result]
With the wayland change to automatically allocate the client side proxy
manually, we can now drop the code (and the FIXME) that did that and just
receive the proxy from the callback arguments.
We don't gain anything from taking a wl_shell_surface in
desktop_surface.set_background, except making wl_shell_surface
gratuitously dependent on wl_shell. In shell.c we can also handle
backgrounds in their own background_configure function which simplifies
the mapping and placement logic.
If the cursor didn't change since last time we had pointer focus we just
wouldn't change it. But whoever had pointer focus in the mean time could
have changed it, so make sure we always set the cursor after pointer enter.
If we can't find a cursor for whatever reason, don't crash the client in
pointer_surface_frame_callback.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Since the introduction of pointer.set_cursor(), it is possible for a
client to set the surface containing the pointer image and get frame
callbacks on it thus allowing a clear implementation of animated
cursors.
This also makes the busy cursor hack of using frame callbacks on the
busy surface unnecessary.
We had duplicated code in many places, using hardcoded paths for
temporary files into more than one path. Some cases did not bother with
O_CLOEXEC, and all hardcoded paths that might not exist.
Add an OS helper function for creating a unique anonymous file with
close-on-exec semantics. The helper uses $XDG_RUNTIME_DIR as the
directory for a file.
This patch unifies the buffer file creation in both Weston and the
clients.
As simple clients are better not linking to libshared, as it would
require e.g. Cairo, they pull the OS compatibility code directly.
Android does not have mkostemp(), so a configure test is added for it,
and a fallback used if it is not available.
Changes in v2:
remove all the alternate possible directory definitions and use
XDG_RUNTIME_DIR only, and fail is it is not set.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
These keymap events communicate the keymap from the compositor to the
clients via fd passing, rather than having the clients separately
compile a map.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Instead of using a uint32_t for state everywhere (except on the wire,
where that's still the call signature), use the new
wl_keyboard_key_state enum, and explicit comparisons.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Instead of using a uint32_t for state everywhere (except on the wire,
where that's still the call signature), use the new
wl_pointer_button_state enum, and explicit comparisons.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
This event lets the compositor inform clients of the canonical keyboard
modifier/group state. Make sure we send it at appropriate moments from
the compositor, and listen for it in clients as well.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Here we create a new client/compositor interface in weston to allow
clients to report their x/y cursor position to the compositor. These
values are then used to center the zoom area on this point. This
is useful for everyone, especially people who are visually impaired.
Add option --self-only to dnd client. If this options is passed, the
drag will be started with no data source so that no drag and drop
events are sent to other clients.
If cairo-gl is used, display_create_surface() will create an
wl_egl_window for each surface and this will result in errors if this
surface is used as a source. Also, one can't get a wl_buffer for such
a surface wich led to crashes when trying to do so for the drag icon.
This patch works around both problems by forcing the item and drag icon
surfaces to use shm.
wl_input_device has been both renamed and split. wl_seat is now a
virtual object representing a group of logically related input devices
with related focus.
It now only generates one event: to let clients know that it has new
capabilities. It takes requests which hand back objects for the
wl_pointer, wl_keyboard and wl_touch interfaces it exposes which all
provide the old input interface, just under different names.
This commit tracks these changes in weston and the clients, as well as
similar renames (e.g. weston_input_device -> weston_seat). Some other
changes were necessary, e.g. renaming the name for the visible mouse
sprite from 'pointer' to 'cursor' so as to not conflict.
For simplicity, every seat is always exposed with all three interfaces,
although this will change as time goes on.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Using the surface enter/leave events track which outputs the window is on and
store those in a "window_output_list" on the window.
To create this list we define a struct window_output that is the list
relationship between the window and the output.
If a client changes it's width/height values in it's widget resize handler,
the input region will be wrong because of the assumptions toytoolkit makes
in frame_resize_handler(). So far, gears is the only client that does this.
A little different from Daniels initial patch. We look up the common
modifiers at xkb init time and convert the xkb serialized modifier mask
to our own modifier bitmask.
To add greater precision when working with transformed surfaces and/or
high-resolution input devices.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Cursor images are fairly small and having one pool for each image adds
a lot of unnecessary overhead. Instead, create one large pool and
allocated all cursor images from that.
In order to do that, however, the code that creates shm surface needed
some refactoring. This patch adds a new struct shm_pool that is used
by the cursor and also changes struct window to use it.
These new protocol events allow us to tell which outputs a surface is on, and
potentially update where we allocate our buffers from.
Signed-off-by: Casey Dahlin <cdahlin@redhat.com>
The biggest performance bottleneck while resizing is the continous
setting up and tearing down of mmaps and faulting in pages. This commit
introduces a per-window pool that we'll allocate buffers out of if it's
available. Then we set initialize it to a big shm pool when we start
resizing and free it when resizing is done.
There was a lot of code here to do a lot of work we didn't need to do.
If we damage a surface with a shm buffer attached, all we need to do
is to re-upload the damaged region to the texture. As for drm buffers,
we don't assume anything changes on attach and only update the
regions the client tells us to update in the damage request.
On one hand, getopt (in particular the -o suboption syntax) sucks on the
server side, and on the client side we would like to avoid the glib
dependency. We can roll out own option parser and solve both problems
and save a few lines of code total.
We just set the input region to the bounding box of the window frame
and set the opaque region to be the opaque rectangle inside the window
if the child widget is opaque.
The X11 compositor currently posts its key presses as keycode - 8; this
is due to X11 having a historical minimum keycode of 8, whereas evdev is
numbered starting from 1. So while the KEY_* constants begin with
KEY_ESC at 1, the corresponding keycode in both X11 and the XKB keymaps
is 9.
window, on the other hand, was relying on xkb->min_key_code being 8 to
translate its keycodes back to useful values in the XKB 'evdev' keycode
map. min_key_code may not always be 8, for restricted subsets of the
keycode map.
Perhaps not the best solution, but at least consistent.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
This change depens on the Wayland core commit:
"protocol: remove absolute coordinates from pointer".
Remove the absolute coordinates from pointer motion and pointer_focus
events.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When window_attach_surface() calls window_get_resize_dx_dy(),
window->resize_edges is cleared. However if there is already a pending
surface to be attached, the resize won't be done until the following
call to window_attach_surface(). In this next call, since resize_edges
is now zero, the top-left corner of the window will be unchanged. If
the user is resizing from the top or left border, this causes the
resize to happen in the wrong direction.
This patch changes window_attach_surface() to call
window_get_resize_dx_dy() only if an attach will actually happen.
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
The code which sends the button events was checking whether there is a
focus widget with a button handler but then always sending the button
event to the grab widget. If the grab widget is different from the
focus widget at this point then it will check the wrong widget for a
button handler and potentially crash. It is also possible for there to
be no grab widget here in the following situation:
1. Press and hold down the left mouse button
2. Press and hold down the right mouse button
3. Release the left mouse button
4. Release the right mouse button
In this case the grab will be released at step 3 because the code only
keeps track of the grab for one button. Then it will try to send the
release event for the right mouse button to a NULL widget so it will
crash.
When a menu self-destructs, free also the widget and struct menu.
As menus are self-destructing, it does not make sense to store the
window pointer, since we cannot clear it automatically. Therefore,
rename window_create_menu() to window_show_menu() that does not return
the window pointer. It also calls window_schedule_redraw() internally.
Fixes Valgrind reported memory leaks.
The alternative would be to explicitly destroy the menu in application's
menu callback.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Input devices may hold a pointer to the widget being destroyed. Reset
such pointers in widget_destroy().
This fixes a use-after-free in window_destroy(), if an application
destroys its widgets before the window.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Fix a memory leak reported by Valgrind, by destroying the window
decorations widget, if it exists.
All widget pointers returned from toytoolkit to the application should
be destroyed by the application explicitly.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
With all input events going to widgets now, we can grab an input device
to a widget, so that all events are delivered to that widgets handlers.
This lets us implement the last bit of the menu behaviour, that is
the client side grabbing of events. The result is that we can now pop down
the menu when we receive clicks in the clients own windows and we
don't send motion and button events to other widgets.
This lands the basic behavior of the popup surface type, but there are still
a number of details to be worked out. Mainly there's a hardcoded timeout
to handle the case of releasing the popup button outside any of the
client windows, which triggers popup_end if it happens after the timeout.
Maybe we just need to add that as an argument, or we could add a new event
that fires in this case to let the client decide whether it ends the popup
or not.
Windows are supposed to be destroyed by the application explicitly.
Deferred tasks are not supposed to be added after returning from
display_run().
Destroy remaining wl objects (except input related will be in a
following patch).
Close the epoll fd.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add a function to destroy the 'struct display', supposedly with all
contained resources, that are not explicitly created by the application.
The implementation at this time is incomplete. It does clean up the
following:
- xkb structure is freed (needs new libxkbcommon from git)
- EGL resources are freed
- wl_display is flushed and destroyed
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add a function, that schedules the display_run() event loop to break
out.
When display_exit() is called, processing continues as usual, until
currently waiting events and deferred tasks have been processed, and
sent requests are flushed. Then, display_run() will return.
This enables toytoolkit apps to handle their exit instead of just being
killed or call exit().
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The Cairo documentation tells us to always call cairo_device_flush()
before using other rendering APIs on the cairo surface, especially where
the Cairo device shares state with us (that is, EGL and GL state in this
case).
Add a call to cairo_device_flush() into display_acquire_window_surface(),
which the toytoolkit offers for switching to native (GL) rendering.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add output_configure_handler as a display property. This exposes only
configured outputs, that is the current mode info is already received,
to applications.
The handler is also called for mode changes on an existing output. This
simplifies the implementation in toytoolkit as we can defer the handler
calls from wl_output binding time to when we receive the current mode.
We do not need separate handlers for "new output" and "mode changed". A
plain "new output" handler would be problematic as the current mode is
not known yet.
Also add delete_handler hook for outputs, but that will never be called
for now, as the protocol lacks a way to signal output disconnections.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Since it is the desktop-shell plugin in the compositor that offers both
wl_shell global interface and wl_shell_surface interface, those are not
available on the tablet-shell plugin.
The tablet-shell client uses the toytoolkit, so toytoolkit must work
somehow even without wl_shell.
Turn all operations in toytoolkit that would require wl_shell or
wl_shell_surface into no-ops.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Change desktop-shell protocol to use wl_shell_surface instead of
wl_surface.
Adapt the desktop-shell client and the shell plugin.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Protocol changes in Wayland core introduced a new interface
wl_shell_surface, and moved all wl_shell surface methods into it. Adapt
the compositor and its Wayland backend, shell plugin, and all clients to
the new interface.
Depends on the Wayland core commit "protocol: introduce wl_shell_surface"
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When a window destroyed, if any input had the window in keyboard
focus, the keyboard focus is reset to NULL. A new keyboard focus is set
only, if the user clicks something. If the user presses a key instead of
clicking, the key press event is sent to the client which has NULL
keyboard focus, triggering a segfault in window_handle_key().
Fix the segfault by ignoring the key event, if there is no target
window.
I triggered this segfault by clicking the unlock dialog away, and then
pressing a key.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Currently, the way to destroy a window in a response to an event (e.g.
button click), is to put a task into the deferred list with
display_defer(). The task will then call window_destroy() from outside
event handling code.
As events are handled, it is possible that the deferred list contains
also the redraw task for this window. As the execution order of these
tasks is unknown (redrawing a freed window is a bug) and redrawing
something that goes away immediately is not useful, the redraw task must
be removed on window_destroy().
'struct input' contains pointers to windows currently in focus for that
input device. These pointers must also be cleared on window_destroy().
This fixes a use-after-free bug for the unlock dialog in desktop-shell
(future commit).
As an irrelevant minor cleanup, window::grab_device member is not used
anywhere, and is removed.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
