compositor-x11: Fix some shutdown crashes

The assertion in x11_compositor_find_output() can trigger during normal
shutdown, for example, when moving the mouse while hitting a hotkey to
close the weston window.

Instead we can remove the assert(), return NULL, and discard events
we can't find a destination output for.

v2 Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
v1 Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

configure: bump libwayland requirements

wl_surface role error codes were added during the 1.6.90 development to
libwayland.

Reported-by: Sjoerd Simons <sjoerd.simons@collabora.co.uk>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: Fix wrong condition to check return value of controller_module_init

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>

text: Don't crash if a seat has no keyboard

A keyboard might not be present in a seat, so check that before
dereferencing keyboard related pointers.

Also, use the keyboard pointer we set to shorten the code a little bit.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Jan Arne Petersen <janarne@gmail.com>

input: swallow the key press event when running a compositor key binding

Reviewed-by: Daniel Stone <daniel.stone@collabora.com>

text: Use bool for input_panel_visible

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>

text: Remove unused linked list from input_method_context

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>

text: Fix coding style

Put the return type on a separate line in function declarations.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>

text: Fix some typos

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>

ivi-shell: rewrite controller API

This fixes the breakage caused by "ivi-shell: make ivi-layout.c as a
part of ivi-shell.so"

hmi-controller.c:
A reference implementation of controller module does not use the layout
functions directly. This get function pointers at controller_module_init
which called by ivi-shell.

ivi-layout-export.h:
Definition of interfaces in a struction: ivi_controller_interface.
function pointers are set at ivi-layout.c.

ivi-layout-private.h:
Definition of some interfaces are remove bucause it is implemented in
ivi-layout.c and set it as function pointer. Several interfaces are kept
here because they are implemented in ivi-layout.c/transition.c but used
in other files.

ivi-layout-transition.c:
remove WL_EXPORT because export is not needed anymore.
remove unnesesary unsed method.

ivi-shell.c:
call controller_module_init of hmi-controller to init it and set
function pointers. If the interface_version is equal or
greater than what hmi_controller was built with, things are pass.
If the interface_version is smaller than what the controller expects,
it has to fail because it cannot work. This is followed the manner,

- never remove a function pointer
- never modify an existing function signature
- additions are allowed at the end of struct ivi_controller_interface
- all function pointers must always be populated and working in
  ivi-shell.so (ivi-layout.c etc.)

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: make ivi-layout.c as a part of ivi-shell.so

ivi-layout.so is separately built and loaded by using dlopen with
RTLD_GLOBAL. This was because these apis defined in ivi-layout.so shall
be used by ivi-modules; e.g. hmi-controller. This shall be improved that
a struct ivi_layout_api contains the whole exported API as function
pointers to be exposed as module_init.

This patch alone builds, but loading controller modules at runtime
failes. This failure will be fixed by following patches.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: several ivi-layout interfaces moved accross source.

Interfaces for properties setting shall be implemented in ivi-layout.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: reordering method in ivi_layout.c

In the future, re-alignmenet of WL_EXPORT per method should be done.
For that work, re-ordering is required to be used by another method.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: clean up unnecesary header

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: clean-up transition animation code.

ivi_layout_transition_layer_render_order is not necesary called. Remove
the method and methods called by it.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: Say something when configuration is missing

When ivi-module is not set in weston.conf, we probably want to tell the
user about their error instead of just dying.

Signed-off-by: Ondřej Majerech <majerech.o@gmail.com>
[Pekka: include Tanibata's suggestion, fix file name, fix surrounding
coding style]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: Propagate errors from module_init

load_modules currently ignores errors signalled by both
weston_load_module and module_init, and instead always returns 0. Its
return value appears to be checked in callers, so we most likely want to
propagate any errors.

Signed-off-by: Ondřej Majerech <majerech.o@gmail.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

clipboard: set fd member of clipboard_source

The fd member of clipboard_source structure was not set
but was used in close().

v2. don't do unnecessary changes

Signed-off-by: Marek Chalupa <mchqwerty@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

screenshooter: fix various memory handling

There were unchecked malloc and no free for this memory.
Also simplify error handling in one function.

v2. remove check if memory is NULL, according to man pages,
    free(NULL) is a no-op

Signed-off-by: Marek Chalupa <mchqwerty@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

window: use roundtrip instead of dispatch after get_registry

wl_display_dispatch() just dispatches events that are in
default and display queues and if there are no events,
then it will wait for them. But only dispatching
the events doesn't guarantee that we got all the global announcements,
we need to do sync too. Therefore use wl_display_roundtrip() instead
of wl_display_dispatch().

Signed-off-by: Marek Chalupa <mchqwerty@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

xwm: fix a crash in handle_state

To produce the bug, build and run: (you don't need the game data to test)

https://github.com/clintbellanger/flare-engine/

$ mv ~/.config/flare ~/.config/flare.bak
$ ./flare  # click 'configure', set full screen mode then click 'ok'
           # weston will crash and dump core.

[Pekka: edited the commit message]
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

tests:  Fix minor typo

compositor: Implement JSON-timeline logging

Logging is activated and deactivated with the debug key binding 't'.
When activated, it creates a new log file, where it records the events.
The log file contains events and detailed object information entries in
JSON format, and is meant to be parsed in sequence from beginning to the
end.

The emitted events are mostly related to the output repaint cycle, like
when repaint begins, is submitted to GPU, and when it completes on a
vblank. This is recorded per-output. Also some per-surface events are
recorded, including when surface damage is flushed.

To reduce the log size, events refer to objects like outputs and
surfaces by id numbers. Detailed object information is emitted only as
needed: on the first object occurrence, and afterwards only if
weston_timeline_object::force_refresh asks for it.

The detailed information for surfaces includes the string returned by
weston_surface::get_label. Therefore it is important to set
weston_timeline_object::force_refresh = 1 whenever the string would
change, so that the new details get recorded.

A rudimentary parser and SVG generator can be found at:
https://github.com/ppaalanen/wesgr

The timeline logs can answer questions including:
- How does the compositor repaint cycle work timing-wise?
- When was the vblank deadline missed?
- What is the latency from surface commit to showing the new content on
  screen?
- How long does it take to process the scenegraph?

v2: weston_surface::get_description renamed to get_label.
v3: reafctor a bit into fprint_quoted_string().

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: add weston_surface_set_label_func()

When printing out logs from Weston's actions, mainly for debugging, it
can be very difficult to identify the different surfaces.  Inspecting
the configure function pointer is not useful, as the configure functions
may live in modules.

Add vfunc get_label to weston_surface, which will produce a short,
human-readable description of the surface, which allows identifying it
better, rather than just looking at the surface size, for instance.

Set the label function from most parts of Weston, to identify cursors and
drag icons, and panels, backgrounds, screensavers and lock surfaces, and
the desktop shell's application surfaces.

v2: renamed 'description' to 'label', so we get
weston_surface_set_label_func().

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: reference implementation of input panel for ivi-shell

- introduces ivi-shell/input-panel-ivi.c which is basically copied
  from desktop shell. It shall be improvaded to remove duplicate
  implementation.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

clients: support ivi-application.xml for clients/window.c

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

clients: support ivi-application.xml for clients/simple-egl.c

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

clients: support ivi-application.xml for clients/simple-shm.c

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: a reference of weston.ini for ivi-shell and ivi-hmi-controller.

- introduces ivi-shell/weston.ini.in

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

data: add reference image files for weston-ivi-shell-user-interface

- introduces reference images used by weston-ivi-shell-user-interface.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

clients: a reference implementation of UI client how to use ivi-hmi-controller.

- introduces ivi-shell-user-interface.c

This is launched from hmi-controller by launch_hmi_client_process and
invoke a
client process.

The basic flow is as followed,
1/ process invoked
2/ read configuration from weston.ini.
3/ draw png file to surface according to configuration of weston.ini
4/ all parts of UI are ready. request "UI_ready" to draw UI.
5/ Enter event loop
6/ If a surface receives touch/pointer event, followings are invoked
according
   to type of event and surface
6-1/ If a surface to launch ivi_application receive touch up, it execs
     ivi-application configured in weston.ini.
6-2/ If a surface to switch layout mode receive touch up, it sends a
request,
     ivi_hmi_controller_switch_mode, to hmi-controller.
6-3/ If a surface to show workspace having launchers, it sends a
request,
     ivi_hmi_controller_home, to hmi-controller.
6-4/ If touch down events happens in workspace,
     ivi_hmi_controller_workspace_control is sent to slide workspace.
     When control finished, event:
ivi_hmi_controller_workspace_end_control
     is received.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: a reference implementation how to use ivi-layout internal APIs

- hmi-controller.so
- introduces hmi-controller.so

The library is used to manage layout of surfaces/layers. Layout change
is triggered by ivi-hmi-controller protocol, ivi-hmi-controller.xml. A
reference how to use the protocol, see ivi-shell-user-interface.c.

In-Vehicle Infotainment system usually manages properties of
surfaces/layers by only a central component which decide where
surfaces/layers shall be. This is differenct use case from desktop
style; each application can request property of its window via xdg-shell
protocol, like fullscreen and set its to top level. In-Vehicle
Infortainment system doesn't allow each application to chagen them from
its application because of safty reasons. The concept of layer is
simillar with a use case of cursor layer of Destop. For In-Vehicle
Infortainment system, it is extended to all applications. For example,
rearview camera application is assigned to a layer to group several
surfaces, e.g. captured image and drawing lines separately. Central
manaegr can control property of the layer of rearview camera.

This reference show examples to implement the central component as a
module of weston.

Default Scene graph of UI is defined in hmi_controller_create. It
consists of
- In the bottom, a base layer to group surfaces of background, panel,
  and buttons
- Next, a application layer to show application surfaces.
- Workspace background layer to show a surface of background image.
- Workspace layer to show launcher to launch application with icons.
  Paths to binary and icon are defined in weston.ini. The width of
  this layer is longer than the size of screen because a workspace
  has several pages and is controlled by motion of input.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

protocol: ivi hmi controller protocol to set up IVI style UI

- introduces ivi-hmi-controller.xml

This protocol realizes following features,
- UI ready
- changing modes; tiling, side by side, full_screen, and random
- Give control a surface; workspace to be controlled by using ivi layout
  APIs
- Display/undisplay a surface; home contains sevaral workspaces to
  launch applications

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: add the shell plugin for In-Vehicle Infotainment system

- ivi-shell.so
- introduces ivi-shell/ivi-shell.[ch]

In-Vehicle Infotainment system traditionally manages surfaces with
global identification. A protocol, ivi_application, supports such a
feature by implementing a request, ivi_application::surface_creation
defined in ivi_application.xml.

The ivi-shell explicitly loads ivi-layout.so and a module to add
business logic like how to layout surfaces by using ivi-layout APIs.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: add IVI layout APIs

- ivi-layout.so
- introduces ivi-layout-export.h, ivi-layout.[ch]

API set of controlling properties of surface and layer which groups
surfaces. An unique ID whose type is integer is required to create
surface and layer. With the unique ID, surface and layer are identified
to control them. The API set consists of APIs to control properties of
surface and layers about followings,

- visibility.
- opacity.
- clipping (x,y,width,height).
- position and size of it to be displayed.
- orientation per 90 degree.
- add or remove surfaces to a layer.
- order of surfaces/layers in layer/screen to be displayed.
- commit to apply property changes.
- notifications of property change.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

protocol: add interface ivi_application extension

- introduces ivi-application.xml

Many applications in an IVI-system are special single-purpose
applications that have a very specific role in the whole IVI UI, for
example a rear camera, speedometer, map, etc. The IVI system vendor
specifies what these are and how they integrate into the UI. They also
vary between particular IVI systems. This is why we use (system-)global,
unique, pre-determined ID numbers to tell what wl_surface is which
application, instead of writing specific shell requests for each one.
Using ID numbers allows vendors to easily invent new component
applications without extending or breaking the actual Wayland protocol.
In IVI-systems, the ID is a standard concept already used in several
APIs, with a vendor-specified global definition of ID assignments.

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

ivi-shell: add README for ivi-shell

Signed-off-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

fbdev: Remove unnecessary semicolon

[daniels: removed extraneous changes]

input: send focus events to the focused client when running a key binding

When running a key binding we don't send the key press to the client
via the wl_keyboard.key event. Instead, send a wl_keyboard.leave/enter
pair so that the client knows the actual state of the keyboard.

Reviewed-by: Daniel Stone <daniel@fooishbar.org>

rpi-renderer: don't pass a box to pixman_region32_contains_point()

We don't care which box contained the point, so don't pass one in.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>

Use zalloc instead of calloc(1, ...)

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Marek Chalupa <mchqwerty@gmail.com>

Setting members to 0/NULL after a zalloc or calloc is redundant

calloc (and zalloc) set the allocated memory to 0, so there's really no
need to do it manually.

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Marek Chalupa <mchqwerty@gmail.com>
[Pekka: dropped the src/evdev.c hunk.]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Use zalloc rather than malloc and manually setting members to 0

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

clients: Use xmalloc in a few more places

Just changes some places where a malloc failure is unhandled
to our xmalloc function that exit()s a little more gracefully.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>

tests: Use one shared xzalloc implementation

Since this is an inlined function, move it to a common header file.

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Marek Chalupa <mchqwerty@gmail.com>

tests: Move ARRAY_LENGTH to a shared header for reuse

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Marek Chalupa <mchqwerty@gmail.com>

tests: Cleanup superfluous headers

Mostly remove headers that aren't actually needed for anything.
Add stdint.h to permit dropping xf86drm.h, which is otherwise unneeded.

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Acked-by: Marek Chalupa <mchqwerty@gmail.com>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

tests: Include "config.h" as first header

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>

dnd: Fix some mouse cursor issues

When ending a drag in the window the cursor will be wrong until the mouse
is moved again.  This is because the item being dragged isn't added
until after the enter event.

Also, when picking up an item while moving the mouse the cursor can switch
back to a non-drag cursor before the drag begins.  This is because of a
slight delay between button click and drag start.

Finally picking up or dropping an item under a second pointer could cause
that pointer to have the wrong cursor.

Closes one of the issues in bug 56298
https://bugs.freedesktop.org/show_bug.cgi?id=56298

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

input: Use slightly less obscure logic in evdev_notify_keyboard_focus()

While the test is actually correct (for non-negative numbers), it's not
at all clear and seems to be an accidental order of operations mistake.

Also, add an assert() to make sure this number is never negative.

Closes bug 86346 - https://bugs.freedesktop.org/show_bug.cgi?id=86346

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Marek Chalupa <mchqwerty@gmail.com>

simple-shm: explain two initial roundtrips

Explain carefully why we need two roundtrips, not just one, not just
dispatch and roundtrip, but two roundtrips after creating the
wl_registry object.

v2: Explain what initial events are, and that this is a general
technique.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Marek Chalupa <mchqwerty@gmail.com>

configure: remove input backend result

Don't print a line with empty value.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: Drop legacy backends in favor of libinput

Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

desktop-shell: don't crash input-panel if no kbd focus

If a keyboard exists but it has no current focus, yet something asks the
input-panel to come up, we would crash here. Check that there is a focus
before attempting to use it.

Maybe there should not even exist a case where input-panel tries to come
up without a keyboard focus, but I am not sure there is no race where it
could happen.

In any case, this fix was brought up by the ivi-shell work, where I
suppose you can somehow hit it.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Cc: Tanibata, Nobuhiko <ntanibata@jp.adit-jv.com>

input: don't run the key bindings on focus in

When getting the focus we get the list of pressed keys, but we are
not supposed to run the key binding on them.

Reviewed-by: Daniel Stone <daniel@fooishbar.org>

tests: Allow tests to use customized command line parameters

Tests will now return the extra command line parameters they need
when run with --params

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: Document options for headless compositor

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
[Pekka: s/Wayland surface/memory surface/, and say default renderer.]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor-headless: add support for transforms set on command line

We now allow --transform with the headless compositor, however it still
doesn't parse anything out of weston.ini

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor-headless: allow rendering with pixman

Allow the headless backend to render into an off screen buffer with pixman.

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

configure.ac: Indicate headless compositor presence in config.h

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

man: explain idle-time=0 in weston.ini

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: Add idle timeout option to weston.ini

Weston's idle timeout can already be set via the '-i' command-line
option, but this patch lets users specify it also via weston.ini.
Note that the command-line option takes precedence over the .ini,
should the option be set by both.

This patch also Updates weston.ini man page with idle-timeout bits

https://bugs.freedesktop.org/show_bug.cgi?id=83921

Signed-off-by: Frederic Plourde <frederic.plourde@collabora.co.uk>
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>

xwayland: fix fullscreening

set_fullscreen has been sending configure before changing the state
and xwayland windows added border to the fullscreen size.

This fixes the bug:
https://bugs.freedesktop.org/show_bug.cgi?id=83502

Signed-off-by: Marek Chalupa <mchqwerty@gmail.com>
Reviewed-by: Ryo Munakata <ryomnktml@gmail.com>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Makefile.am: Link weston_multi_resource against rt library

This fixes this build failure:

/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.3/../../../../x86_64-unknown-linux-gnu/bin/ld:
clients/weston_multi_resource-multi-resource.o: undefined reference to
symbol 'clock_gettime@@GLIBC_2.2.5'
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.3/../../../../x86_64-unknown-linux-gnu/bin/ld:
note: 'clock_gettime@@GLIBC_2.2.5' is defined in DSO /lib64/librt.so.1
so try adding it to the linker command line
/lib64/librt.so.1: could not read symbols: Invalid operation

gl-renderer: compress pixman bands to simplify geometry

Pixman uses y-x banded rectangles to represent regions.  We use these
y-x banded rectangles to generate triangle fans, resulting in more
geometry than strictly necessary to draw the screen.

This patch combines the bands to reduce geometry for complex scenes.

Acked-by: "Jasper St. Pierre" <jstpierre@mecheye.net>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

backends: refactor transform string parsing

Most of the backends do their own parsing of transform strings, so let's
put that all in the same place (compositor.c/h)

Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

cairo-util: fix shadows for small clients

If the client is small (< 128 pixels in any ward),
then the shadows overlap and create dark lines behind clients.
This is a problem mosly with pop-up menues. The lines become observable
when the menu has less than three items. The other case is when
the client doesn't restrict its size when resizing (try
'weston-eventdemo --max-width=1 --max-height=1' for example)

This fixes a part of the bug:
https://bugs.freedesktop.org/show_bug.cgi?id=78511

v2:
  - rework computing of the size of corners
  - rewrite some comments
  - rename tile_mask to render_shadow (in separate patch)

Signed-off-by: Marek Chalupa <mchqwerty@gmail.com>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

toytoolkit: rename tile_mask to render_shadow

This function is used and clearly designed only for drawing the shadows.
Rename it so that it has name after what it does and also move some
common code into the function.

Signed-off-by: Marek Chalupa <mchqwerty@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

protocol: Fix some typos in protocol files

Signed-off-by: Seedo Eldho Paul <seedoeldhopaul@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

Revert "input: don't send to clients key events eaten by bindings"

This reverts commit 5c11fc6fb7d9b1bccd50fc982c0798c9e09a2a04.

According to two input specialists, this was the wrong way:
http://lists.freedesktop.org/archives/wayland-devel/2014-November/018287.html

Cc: Peter Hutterer <peter.hutterer@who-t.net>
Cc: Daniel Stone <daniel@fooishbar.org>
Cc: Giulio Camuffo <giuliocamuffo@gmail.com>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: do not overwrite error exit code

If there are several things that cause an exit, keep the error code from
the first one, not the last one. The latter ones could be just fallout
from the first.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: Return a user-defined exit code

Currently, once we've reached our main's wl_display_run(), we always
return ret=EXIT_SUCCESS when weston terminates through wl_display_terminate.

This patch makes it possible to specify another return value by setting
prior to terminating Weston. This is useful for automated tests that want
to report test failures to the overlying testing infrastructure.

Signed-off-by: Frederic Plourde <frederic.plourde@collabora.co.uk>
[Pekka: fixed some tabs.]
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

xwm: find a seat for all the surface's views

This allows to move or resize a xwayland client by acting on all
its views.

input: use tmp var in pointer_unmap_sprite

Remove lots of repeated indirection.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

input: don't send to clients key events eaten by bindings

weston key bindings are supposed to eat the key events, and not pass it
on to clients, and indeed the wl_keyboard.key event is not sent. But
we must also not put the key in the keys array to pass to client with
the wl_keyboard.enter event, or else we may send the 'eaten' one too.
In the case of a key binding hiding a surface having the keyboard focus,
the shell may decide to give the focus to another surface, but that will
happen before the key is released, so the new focus surface will receive
the code of the bound key in the wl_keyboard.enter array.

Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: add missing newline to help text

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: List only available backends in help text

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

compositor: Display option help text only for enabled backends

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

compositor: Fix typos in help text

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

compositor: Fix tabbing of help output

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

compositor: (Whitespace-only) Indent with tabs

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>

man: Fix some weston.ini.man typesetting macros

This patch fixes a couple of misuses around .TP 7 macros that wrongly
limited right margins and relative identation.

Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

smoke: fix valgrind invalid read errors

There are a number of invalid read errors reported by valgrind of the
form:
==13428== Invalid read of size 4
        ==13428==    at 0x405656: advect (smoke.c:116)
        ==13428==    by 0x405E80: redraw_handler (smoke.c:228)
        ==13428==    by 0x40DE74: widget_redraw (window.c:3995)
        ==13428==    by 0x40E02D: surface_redraw (window.c:4053)
        ==13428==    by 0x40E0C9: idle_redraw (window.c:4082)
        ==13428==    by 0x410FC9: display_run (window.c:5561)
        ==13428==    by 0x406518: main (smoke.c:373)
        ==13428==  Address 0xb2c9b14 is 4 bytes after a block of size
                   160,000 alloc'd
        ==13428==    at 0x4C29DB4: calloc
        ==13428==    by 0x40646B: main (smoke.c:360)

This results in invalid rendering when running a debug version of the
application.

Fix the issue by limiting the maximum values of px and py to 1.5 less
than width and height. This prevents reading past the end of the source
buffer.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=82287
Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

smoke: Don't commit an xdg_surface with a NULL buffer

Committing to an xdg_surface with a NULL buffer is currently illegal in
the mutter implementation, so this simply causes the client to error and
exit.

It seems the reason the client did this was so it could add its own
frame callback, but toytoolkit actually provides accurate everything we
need. Just use its functions instead to get the time and schedule a
redraw.

Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
Reviewed-by: Giulio Camuffo <giuliocamuffo@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

input: fix access to invalid resource

the keyboard focus surface may not have a valid resource (server side
surface or a surface surviving its client), so check if it is valid
before using it.

Acked-by: Marek Chalupa <mchqwerty@gmail.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: make keep_buffer a bool

The comments already call it bool, so let's just make it one

Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: stop sending spurious events on mode switch

commit 6ae7bc982fa accidentally made weston_output_mode_switch_temporary
send done events when it shouldn't.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: check version before calling wl_output_send_scale()

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

input: allow to destroy bindings in their handler

use wl_list_for_each_safe to iterate on the bindings list when
firing them, this way a binding can safely be destroyed in its
function handler.

Acked-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: Refactor weston_output_mode_switch()

This breaks weston_output_mode_switch() into 3 functions:
weston_output_mode_set_native()
weston_output_mode_switch_to_temporary()
weston_output_mode_switch_to_native()

Differences from previous behaviour:
SET_NATIVE didn't set current_scale (now it does)
SET_TEMPORARY could set mode and scale independently - now it can't.

Signed-off-by: Derek Foreman <derekf@osg.samsung.com>
Pekka Paalanen <pekka.paalanen@collabora.co.uk>

compositor: Fix weston_subsurface_is_synchronized() return value.

Commit 280e7dd918f1717c7d677676384a9cd991097741 introduced a bug in the
return value of weston_subsurface_is_synchronized().

Signed-off-by: Carlos Olmedo Escobar <carlos.olmedo.e@gmail.com>
Reviewed-by: Derek Foreman <derekf@osg.samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

window : compare version and call appropriate destructor

[Pekka Paalanen: removed trailing space, split long line.]

Signed-off-by: kabeer khan <kabeer.khan@samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

data_device : change version while initializing data_device_manager interface and data_device interface

[Pekka Paalanen: split long lines, dropped MIN().]

Signed-off-by: kabeer khan <kabeer.khan@samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

cosmetic: replace boolean function return values with bool

For functions that test if something is true/valid and return a 1
or 0, it makes sense to switch to bool.

Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>

gl-renderer: don't move memory in output_rotate_damage

output_rotate_damage shifted an array of pixman regions with a loop.  Now
it uses an index into that array.

This currently only saves 1 pixman_region32_copy, but we can now raise
BUFFER_DAMAGE_COUNT without a performance impact if we run into a
configuration where this is useful.

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

Very OCD cosmetic nitpick

releasing: update web host name

Signed-off-by: Daniel Stone <daniels@collabora.com>

tests: add role conflict sub-surface vs. wl_shell_surface

Add tests for triggering the role conflict when a wl_surface is already
a wl_shell_surface and then attempted to be made into a sub-surface, and
vice versa.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Jasper St. Pierre <jstpierre@mecheye.net>

compositor: send error for surface role resets

With the more accurate definition of wl_surface roles in Wayland,
enforce the restriction: a role is always set permanently, and
attempting to change it is a protocol error.

This patch is based on Jasper's patch:
http://lists.freedesktop.org/archives/wayland-devel/2014-August/016811.html

The difference in this patch compared to his are:

- send role errors on the interface whose request triggers it, not on
  wl_surface

- an interface could have several requests assigning different roles,
  cannot use wl_interface as the unique key; use an arbitary string
  instead

- ensure in window-manager.c that create_shell_surface() ->
  create_common_surface() is never called with surface->configure set,
  to avoid compositor abort

- use wl_resource_post_no_memory() where appropriate instead of
  hand-rolling it with wl_resource_post_error()

Ideally we would not add weston_surface::role_name field, but use
weston_surface::configure. At the moment this is not possible though,
because at least shell.c uses several different roles with the same
configure function. Drag'n'drop uses two configure functions for the
same role. The configure hook is also reset in several places,
which is not good for role tracking.

This patch overlooks the wl_surface roles assigned in privileged
extensions: screensaver, panel, background, lock, input panel.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Jasper St. Pierre <jstpierre@mecheye.net>

protocol: add role error code to xdg_shell

Needed for properly reporting role violations from
xdg_shell.get_xdg_surface and .get_xdg_popup.

Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Reviewed-by: Jasper St. Pierre <jstpierre@mecheye.net>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>