With the introduction of the use of buffer scaling in ed4fcee4ct we
must request version 3 of the compositor as that is the version of the
surface interface that adds this new functionality. See the following
commit in weston:
commit a85118c1b85df6fbf8f896dca971a5b79a94da71
Author: Jason Ekstrand <jason@jlekstrand.net>
Date: Thu Jun 27 20:17:02 2013 -0500
Use wl_resource_create() for creating resources
This commit sets the version numbers for all added/created objects. The
wl_compositor.create_surface implementation was altered to create a surface
with the same version as the underlying wl_compositor. Since no other
"child interfaces" have version greater than 1, they were all hard-coded to
version 1.
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
https://bugzilla.gnome.org/show_bug.cgi?id=703817
Ths allows the retrieval of the wl_surface before the window is shown.
The surface is still created in the original places since the surface
and shell surface is destroyed when the surface is programmatically
hidden.
We've long had double precision mouse coordinates on wayland (e.g.
when rotating a window) but with the new scaling we even have it on
X (and, its also in Xinput2), so convert all the internal mouse/device
position getters to use doubles and add new accessors for the
public APIs that take doubles instead of ints.
This lets use use a scaled Xft/DPI for old apps while not
blowing up the size of scaled windows. Only apps supporting
Gdk/WindowScaleFactor should supprt Gdk/UnscaledDPI.
If you set GDK_SCALE=2 in the environment then all windows will be
scaled by 2. Its not an ideal solution as it doesn't handle
multi-monitors at different scales, and only affects gtk apps.
But it is a good starting points and will help a lot on HiDPI
laptops.
We track the list of outputs each window is on, and set the
scale to the largest scale value of the outputs. Any time the scale
changes we also emit a configure event.
We bind to the newer version of the wl_output which supports
the new done and scale events, and if we use this to get the
scale for each monitor (defaulting to 1 if not supported).
If a cairo_surface for a window has a device scale set we need
to respect this when creating a similar window. I.e. we want
to then automatically create a larger window which inherits
the scale from the original.
We also need to calculate a different device_offset if there
is a device_scale set.
-Don't include unistd.h unconditionally as it's not available in Visual
Studio, but include io.h where necessary.
-Avoid C99isms, and use _chsize_s in place of ftruncate when unistd.h is
not available (as in the case of Visual Studio)
Only look at "impl" windows in gdk_window_process_updates_with_mode()
since these are the only ones we care about. This avoids a lot of
unnecessary calls to g_list_copy() and g_object_ref().
We don't want to recurse into children that are clipped, as that is
wasted work. We handle this by moving the empty check to the top
of the function and only using the clipped region everywhere.
Move the call to gdk_x11_atom_to_xatom_for_display() outside of the
search loop in gdk_x11_screen_supports_net_wm_hint(). In my test case
(running Audacious for about a minute), this reduced the total number of
hash table lookups performed from 370,000 to 230,000.
https://bugzilla.gnome.org/show_bug.cgi?id=702913
and not on each keystroke, which for some IMs apparently caused a full
update on each keystroke, not just a check for changes. Patch from
Takuro Ashie, bug #698183.
window_type_hint_to_level(): applied patch from Paul Davis which moves
dialogs to NSFloatingWindowLevel. This is not quite the perfect
solution, but it's a pragmatic fix that makes apps which have both
window types much more usable, and prevents dialog from disappearing
under an application's main window.
(cherry picked from commit 59d49e1566)
Include config.h first so that _GDK_EXTERN may be defined once
and only once during the build, so that we do not get warnings/
errors for macro redefinition.
https://bugzilla.gnome.org/show_bug.cgi?id=701251
Include gdkwin32.h (which includes gdkprivate-win32.h and gdkwin32cursor.h
during the build of GDK-Win32) so that
gdk_win32_icon_to_pixbuf_libgtk_only() and
gdk_win32_pixbuf_to_hicon_libgtk_only() get exported, so that the GTK
DLL can link correctly.
Calling XIQueryPointer() on a slave device is going to trigger a
BadDevice X error. So in case we query a slave device state, ask the
master device instead.
https://bugzilla.gnome.org/show_bug.cgi?id=700233
Add missing check in gdk_x11_device_manager_xi2_get_window(), returning
NULL if no valid XI2Event* has been found. Calling code seems to be
prepared to handle NULLs coming from this function, so it should be
safe enough (e.g. check gdk_event_source_get_filter_window()).
https://bugzilla.gnome.org/show_bug.cgi?id=700465
We don't track the full clip for each window anymore, as this
is not useful when no windows are opaque. However, we still
need the full clip for the shape, so its calculated manually.
However, it was previously only recalculated when the clip changes
which doesn't correctly handle the case of a sibling geometry changing.
So, instead of doing this directly when geometry changes we just
set a bit in the toplevel whenever some window geometry changes, and
we then handle this in process_updates, updating the shape for all
native windows. This should be ok performance-wise because we don't
expect a lot of native children.
This is left over from an old effort to bundle gtk and its dependencies
into a Mac OS X Framework. The effort was more or less successful but
proved difficult to maintain and impractical to use because gtk programs
don't use the special Framework include syntax.
In the ancient X days you could have Xservers that had multiple active windows, like
one truecolor and one 8bit palette. Then most apps ran in 8bpp but a single window
would use truecolor. This is done by specifying different visuals for the windows.
To make this work we ensured that a window with a visual different from its parent
gets a native subwindow, so that X can tell the hardware to do its magic.
These days the only real time we get two different visual is when one is a rgba visual
and the other is not. So, the code to check this doesn't really do anything but
get in the way when someone accidentally manages to not get a rgba visual on
a child window (see bb7054b508). So, to avoid
such errors we just remove the "different visual than parent" check.
We need to send exposes for all native windows, even the ones
without an exposure mask set, because otherwise non-native
children of the native window with an exposure mask will
not be drawn.
This removes the typechecks in GDK_WINDOW_TYPE and GDK_WINDOW_DESTROYED. These
are only used internally in gdkwindow.c and gdkdisplay.c anyway, and these
functions check for typesafety of arguments on function entry.
This function returns all the children that has a specific user_data set.
This is used a lot in the new GtkWidget drawing code and doing
it this way is faster than getting every child and calling get_user_data
on each (which was a non-neglible part of the profiles). Additionally it
also allows use to use some kind of hashtable to make this operation even
faster if needed in the future.
This lets you register callbacks for when child widgets invalidate
areas of the window read it and/or change it.
For instance, this lets you do rendering effects and keeping offscreen
caches uptodate.
If we got the release event for the last buffer then we're
fine with writing directly to the window surface, as wayland
will not be looing at it. This saves us from allocating
and copying more data.
First of all, we now only do paints on native windows, as there is
really no reason anymore to do it for subwindows. Secondly, we
keep track of the paints even for GtkPaintable windows, but for
that case we don't create the offscreen surface, but rather
assume the windowing system does the backing store.
In the case where the layout phase queued a layout we don't
want to progress to the paint phase with invalid allocations, so
we loop the layout. This shouldn't normally happen, but it may
happen in some edge cases like if user/wm resizes clash with
natural window size changes from a gtk widget. This should not
generally loop though, so we detect this after 4 cycles and
print a warning.
This was detected because of an issue in GtkWindow where it
seems to incorrectly handle the case of a user interactive resize.
It seems gtk_window_move_resize() believes that configure_request_size_changed
changed due to hitting some corner case so it calls
gtk_widget_queue_resize_no_redraw(), marking the window as need_alloc
after the layout phase. This commit fixes the issue, but we should
also look into if we can fix that.
Now that all windows are non-opaque we can simplify the invalidation
a lot. There is no need to clip the invalidate area to child regions,
because we will always redraw everything under all the children.
We only have to handle native childen specially.
We now only do one expose event per native window, so there will
only be one begin/end_paint() call. This means all the work with
implicit paints to combine the paints on a single double buffer
surface is unnecessary, so we can just delete it.
We now consider non-native windows non-opaque, which means any invalid
area in a subwindow will also be invalid all the way up to the nearest
native windows. We take advantage of this by ignoring all expose events
on non-native windows (which typically means just the toplevel) and instead
propagating down the draw() calls to children directly via
gtk_container_propagate_draw.
This is nice as it means we always draw widgets the same way, and it
will let us do some interesting ways in the future.
We also clean up the GtkWidget opacity handling as we can now always
rely on the draing happening via cairo.
We can't really just draw by walking down the widget hierarchy, as
this doesn't get the clipping right (so e.g. widgets doing cairo_paint
may draw outside the expected gdkwindow subarea) nor does it let
us paint window backgrounds.
So, we now do multiple draws for each widget, once for each GdkWindow,
although we still do it on the same base cairo_t that we get for the
toplevel native window. The difference is only the clipping, the rendering
order, and which other widgets we propagate into.
We also collect all the windows of a widget so we can expose them inside
the same opacity group if needed.
NOTE: This change neuters gtk_widget_set_double_buffered for
widgets without native windows. Its impossible to disable
the double buffering in this model.
Since we dropped the move region optimization there is really no need
to try carefully keep track of opaque non-overlapped regions, as we
don't use this information to trigger the optimization anymore.
So, by assuming that all windows are non-opaque we can vastly simplify
the clip region stuff. First of all, we don't need clip_region_with_children,
as each window will need to draw under all children anyway. Secondly, we
don't remove overlapping sibling areas from clip_region, as these are
all non-opaque anyway and we need to draw under them
Finally, we don't need to track the layered region anymore as its
essentially unused. The few times something like it is needed we can
compute it explicitly.
For the case of native children of widgets we may cause a repaint
under native windows that are guaranteed to be opaque, but these
will be clipped by the native child anyway.
This basically neuters gdk_window_move_region, gdk_window_scroll
and gdk_window_move_resize, in that they now never copy any bits but
just invalidate the source and destination regions. This is a performance
loss, but the hope is that the simplifications it later allows will let
us recover this performance loss (which mainly affects scrolling).
Change the visibility handling to be the same way we do it in
GLib now. We pass -fvisibility=hidden to gcc and decorate public
functions with __attribute__((visibility("default"))).
This commit just does this for GDK, GTK+ will follow later.
Make all GDK_DEPRECATED and GDK_AVAILABLE macros use a
new _GDK_EXTERN macro. _GDK_EXTERN defaults to just 'extern'
but a subsequent commit will add visibility handling to it
while building GTK+.
glib-mkenums is not currently clever enough to know which version an
enum type was added in, so just mark all the _get_type() functions as
available in all versions.
Add a macro to declare that a particular symbol is available in all
versions of GTK+.
All newly-added symbols should have proper version macros (like
GDK_AVAILABLE_IN_3_4).
When we call _gdk_wayland_display_load_cursor_theme during
the initial opening of the first display, gdk_setting_get does
not work yet, since it relies on the default display/screen
being set, which only happens after open returns.
Instead, just use the screen of this display.
It's not necessary anymore because gdk_display_manager_get() always
succeeds and the value is independant of when it was called as it's no
longer backend specific.
Move it from GdkDisplayManagerX11.init to GdkDisplay.class_init.
This shouldn't cause any problems, but who knows, so keep this patch
small.
Reason for this is the unification of display managers.
There is currently no Wayland protocol for providing presentation
timestamps or hints about when drawing will be presented onscreen.
However, by assuming the straightforward algorithm used by the
DRM backend to Weston, we can reverse engineer the right values.
https://bugzilla.gnome.org/show_bug.cgi?id=698864
Combine duplicate code for creating and destroying surfaces.
To make the operation of the destroy() operation more obvious, the
destruction of the (fake) root window at display dispose time is
changed to not be a "foreign" destroy.
https://bugzilla.gnome.org/show_bug.cgi?id=698864
Use wl_surface_frame() to get notification when the compositor paints
a frame, and use this to throttle drawing to the compositor's refresh
cycle.
https://bugzilla.gnome.org/show_bug.cgi?id=698864
Lazily creating the cairo surface that backs a window when we
first paint to it means that the call to
gdk_wayland_window_attach_image() in
gdk_wayland_window_process_updates_recurse() wasn't working the
first time a window was painted.
https://bugzilla.gnome.org/show_bug.cgi?id=698864
When exposing an area, we were individually damaging and committing
each rectangle, *before* drawing. Surprisingly, this almost worked.
Order things right and only commit once.
https://bugzilla.gnome.org/show_bug.cgi?id=698864
This commit is very similar to 8c8853a1f5
We update the keynames.txt file from gdkkeynames.h, and we update
keynames-translate.txt to include all the keysym names that we want
to have translations for. Also strip the XF86 from the translatable
keysym names, since we are returning those names now from
gdk_keyval_name().
keyname-table.h is regenerated from these updated files.
Instead of GdkDisplay::init, only add the display to the display manager
in GdkDisplay::opened. This avoids spurious changes of the default
display in gtk_init() when we're trying to find the one that works and
try to open lots of different ones.
This makes Wayland and X11 no longer call into XKB and libX11 for these
functions but use GDK's own copy of these functions, just like the
win32, quartz and broadway backends.
A function was doing nothing but calling a function that was in its own
source file doing nothing but calling a function in its own source file
that did nothing.
This is another step towards making GdkDisplayManager backend-agnostic.
Most of the backends profit from this as their atom implementations
where generic anyway - x11 needed that to allow multiple X displays and
broadway, quartz and wayland don't have the concept of displays.
The X11 backend still did things, so I only #if 0'd some code but did
not actually update anything.
In the case that the client is started directly by the compositor the
WAYLAND_SOCKET environment variable is set containing the fd to use that was
created by a socketpair.
This environment variable is consumed by a call to wl_display_connect so a
second call will not take advantage of it.
https://bugzilla.gnome.org/show_bug.cgi?id=697673
If gdk_window_flush_outstanding_moves() creates new update area
we handle this directly in the same draw to avoid flashing.
This mainly affects win32 as X11 does its exposes from moves async.
However, its important for win32 since ScrollDC seems to sometimes
invalidate (and not copy) unexected regions.
http://bugzilla.gnome.org/show_bug.cgi?674051
Rather than set the window update region and repaint this region
when we get a WM_PAINT we just directly add it to the update
region. No need to roundtrip via win32.
This lets us also make sure we do this drawing in the same update
cycle. This seems especially important on Win7, because ScrollDC
seems to act kind of weird there, not using bitblt in areas where
it seemingly could, which makes scrolling look really flashy.
http://bugzilla.gnome.org/show_bug-cgi?id=674051
On crossing events resulting from moving windows (eg. workspace switch),
deviceid equals sourceid, so make those reset scroll valuators on all
slave devices to avoid misleading jumps in scroll events
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=690275
Under Wayland we don't know the absolute position of the device but there are
some API calls that expect to get an root window position. Previously we were
not assigning any value to these out parameters potentially leaving the values
undefined.
This change returns the current surface relative position of the device.
The is_modifier field is supposed to be set if the key
would act as a modifier, not if any modifiers are currently
active. To fix this, introduce a private
_gdk_wayland_keymap_key_is_modifier function.
At the same time, make the hardware_keycode field in key
events actually contain the hardware keycode, not a copy
of the keyval.
We always emit direction-changed when we get a new keymap, but
for state changes, we compare old and new direction and only
emit the signal when the direction actually changes.
We can get G_IO_HUP and G_IO_IN at the same time, if the compositor writes
data to us and then closes our connection. Make sure that we dispatch events
always if we have G_IO_IN and then error out if we get G_IO_HUP after that.
This way we don't have to reopen all the time for pure updates,
and we can immediately unlink the shm file to avoid "leaking" them
on improper shutdown.
We now only update surface data after we have painted. Before we painted
in an idle, which meant we might send black data some times if we e.g.
resized the window and had not painted yet. Also, it means we're updating
less often to the daemon, saving resources.
We still have to queue a flush in the idle for non-draw operations,
otherwise e.g. resize of a toplevel will never be flushed if the clock
is frozen (e.g. during toplevel resize).
We don't want to update the window size on configure event, only
the position, as the size is client side controlled. We were
updating to an old size during resizes which causes us to send
surfaces of the wrong size to the daemon.
The cursor buffer is only non-null when a cursor is created from pixbuf,
so it is not necessary to keep track of whether to free this buffer on
finalize.
By keeping a pointer to the wl_cursor struct in GdkWaylandCursor, it is
no longer necessary to duplicate cursor data (width, height, hotspots,
etc.) between wl_cursor and GdkWaylandCursor.
Instead of maintaining the init refcount in regular event handlers that can
fire in case of hotplug or mode changes, use a dedicated sync callback
to wait for roundtrips.
If a window is overlapped by a layered (i.e. partially transparent)
window then that region will not disappear from the native window clip
region. This lets us handle compositing multiple layers of windows.
For native subwindows this doesn't really work. For them we apply the
clip region as a shape to the native window which lets us have client
side windows overlapping the native window. However, with the addition
of the layered stuff the "overlapped-by-alpha-csw" part got broken, as
this area is not removed from the clip region of the native window.
We fix this by also removing the layered area when applying the shape.
This means alpha and alpha backgrounds don't work over native windows,
but there is not much to do about that.
https://bugzilla.gnome.org/show_bug.cgi?id=696370
The global_removal argument is the _name_ of the object.
We were comparing it to the _object id_ of the object.
To fix this, store the name at the time the object is bound.
We need to be a bit more careful when updating the screen
size - the code that was there would not do the right thing
if e.g. the width of one monitor was reduced.
We use a ref-count mechanism to track whether parts of the init sequence
still needs round trips to receive remaining initial state. Typically
we need a couple of roundtrips total to get the global list, then the
input and output configurations, but with the ref-count we avoid making
global assumptions like that.
https://bugzilla.gnome.org/show_bug.cgi?id=696340
Allow to set a GdkWindow to use a custom surface instead of a
wl_shell_surface. It allows to register the surface as a custom type
with some Wayland interface.
https://bugzilla.gnome.org/show_bug.cgi?id=695861
The GDK model for keymaps expects the keymap object to stay
around and emit a ::keys-changed signal. So, do that. This
should make layout changes work, but it remains untested since
weston does not support layout changes at runtime.
At the same time, plug a memory leak where GdkWaylandKeymap
forgot to free its xkb objects in finalize.
https://bugzilla.gnome.org/show_bug.cgi?id=696339
For introspection scanning which ends up calling class_init() which in
turn calls into the keybindings code, we can just use the fallback
keyval conversion code.
https://bugzilla.gnome.org/show_bug.cgi?id=696457
With this commit, we pick up xft settings from GSettings
as well. Among other things, this makes the Large Text
setting work. Still to do: pick up fontconfig changes without
having all clients use up inotify watches for all font
directories.
Add GInitable interface with a default implementation that always
succeeds. This allows backends to override the GInitable implementation
and add their own checks to determine if the backend can be loaded. If
a backend cannot be loaded, GDK can attempt to load the next available
backend.
Since backends may need to read any relevant options (such as the
display flag) to determine if they can be created successfully, this
patch also removes calls that attempt to create the display manager
before the options have been parsed.
https://bugzilla.gnome.org/show_bug.cgi?id=694465
The check for GDK_CURSOR_IS_PIXMAP was ineffective, since _all_
cursors have this type, from the looks of it. Instead, store
buffer ownership information separately.
These might be candidates for a future settings interface; until
then, we use GSettings directly. Note again that we are careful
to avoid a dependency on GNOME schemas.
Key repeat under X is not affected by modifiers. And on some systems
(e.g my Thinkpad), NumLock is permanently on, rendering key repeat
nonfunctional. This commit changes the Wayland backend to do
key repeat regardless of modifiers.
http://bugzilla.gnome.org/show_bug.cgi?id=695497
Commit 0d9d808217 fixed the hotspot issue,
but commit f2cc52fddd then optimized away
cursor changes a little too aggressively. We always need to set the
cursor on enter. Make sure we clear the current cursor on leave so we
don't think it's already set on the next enter.
https://bugzilla.gnome.org/show_bug.cgi?id=695512
Until we figure out where we want to go with settings under
Wayland, this makes GTK+ applications a lot easier to deal
with under Wayland.
Note that we are careful to deal with the absence of schemas,
so this does not introduce a dependency on GNOME settings.
wl_pointer.set_cursor is rejected if the serial number doesn't match
the enter serial number for the wl_pointer. We passed the right serial
number when setting the cursor surface in response to the enter event.
Later set_cursor requests fail, but we can still attach new buffers to
our cursor surface, which is why the cursor changed, but the hotspot
didn't update. Clicking in the decoration results in a leave/enter pair
which triggers wl_pointer.set_cursor with the right serial. That's why
clicking the decoration sets the right cursor.
https://bugzilla.gnome.org/show_bug.cgi?id=695512
We need to pass the delta between the old and new hotspot
when attaching the new cursor surface, to keep the hotspot
at the same position. We can't deal with this in the compositor,
since the set_cursor call already overwrites the old hotspot,
so the information is lost by the time the attach happens.
Unfortunately, we can't query the initial hotspot from
the compositor, so the first cursor change will make the
hotspot jump.
https://bugzilla.gnome.org/show_bug.cgi?id=695512
Use separate fields for saving the window dimensions prior to fullscreening
and maximisation. Then use those fields to restore the window dimensions from.
Requests are not limited in size by BroadwayRequest, as
BroadwayRequestTranslation can be of variable size. No need
to copy the request anymore though, because requests are aligned
now.
With recent changes in attach semantics, we always need to attach before
committing. Without this changes to the window contents to not get reflected
in the content of the surface.
Signed-off-by: Rob Bradford <rob@linux.intel.com>
We currently use this information to display the title
string in the window list of the desktop shell.
Signed-off-by: Rob Bradford <rob@linux.intel.com>
Both of them started to make use of round(), a C99 function. So, include
fallback-c89.c to provide a fallback implementation for round() for
compilers that don't have round()
https://bugzilla.gnome.org/show_bug.cgi?id=694339
If no updates, redraws, or repaints have been scheduled for this frame,
we will skip immediately to RESUME_EVENTS, and no GdkFrameTimings will
be created.
https://bugzilla.gnome.org/show_bug.cgi?id=694732
When events are paused, we should not return TRUE from prepare() or check().
GTK+ handles this for events that are already in the GTK+ queue, but
we also need suppress checks for events that are in the system queue - if we
return TRUE indicating that there are events in the system queue, then we'll
call dispatch(), and do nothing. The event source will spin, and will never
run the other phases of the paint clock.
(Broadway doesn't have a window system queue separate from the GDK event queue,
but we write the function the same way for consistency.)
https://bugzilla.gnome.org/show_bug.cgi?id=694274
When events are paused, we should not return TRUE from prepare() or check().
GTK+ handles this for events that are already in the GTK+ queue, but
we also need suppress checks for events that are in the system queue - if we
return TRUE indicating that there are events in the system queue, then we'll
call dispatch(), and do nothing. The event source will spin, and will never
run the other phases of the paint clock.
https://bugzilla.gnome.org/show_bug.cgi?id=694274
Commit 1db87c897f accidentally removed
a check for !in_paint_idle in maybe_start_idle which causes us
to create a paint loop whenever something requests a phase
inside the paint_idle.
The default windows timer resolution is 16msec, which is too little
for fluent animations (say at 60Hz). So, while a paint clock is
active we temporarily raise the timer resolution to 1 msec.
Add an API to start or stop continually updating the frame clock.
This is a slight convenience for applcations and avoids the problem
of getting one more frame run after an animation stops, but the
primary motivation for this is because it looks like we might have
to use timeBeginPeriod()/timeEndPeriod() on Windows to get reasonably
accurate timing, and for that we'll need to know if there is an
animation running.
https://bugzilla.gnome.org/show_bug.cgi?id=693934
gdkwindown-win32.c included windows.h directly rather than via gdkwin32.h
which broke the build for me at least. Instead rely on it being included in
gdkwin32.h and things work right.
The macros we had for checking for toplevel windows were passing
through the root window, which was not intentional and meant that
for the root window WINDOW_IS_TOPLEVEL() returned TRUE but
window->impl->toplevel was NULL, causing gdk_window_create_cairo_surface()
to crash.
A switch of device may be significant for an application, so don't
compress motion events if they are for different devices. This simple
handling isn't sufficient if we have competing event streams from
two different pointer events, but we don't expect this case to be
common.
* remove gdk_frame_clock_get_frame_time_val(); a convenience
function that would rarely be used.
* remove gdk_frame_clock_get_requested() and
::frame-requested signal; while we might want to eventually
be able to track the requested phases for a clock, we don't
have a current use case.
* Make gdk_frame_clock_freeze/thaw() private: they are only
used within GTK+ and have complex semantics.
* Remove gdk_frame_clock_get_last_complete(). Another convenience
function that I don't have a current use case for.
* Rename:
gdk_frame_clock_get_start() => gdk_frame_clock_get_history_start()
gdk_frame_clocK_get_current_frame_timings() => gdk_frame_clock_get_timings()
Since we're not exporting the ability to create your own frame
clock for now, remove the setters for GdkFrameTimings fields.
Also remove all setters and getters for fields that are more
about implementation than about quantities that are meaningful
to the applcation and just access the fields directly within
GDK.
Now that GdkFrameClock is a class, not interface, there's no real advantage
to splitting the frame history into an aggregate object, so directly
merge it into GdkFrameClock.
It's unlikely that anyone will want to have, say, a GtkWidget that
also acts as a GdkFrameClock, so an abstract base class is as
flexible as making GdkFrameClock an interface, but has advantages:
- If we decide to never make implementing your own frame clock
possible, we can remove the virtualization.
- We can put functionality like history into the base class.
- Avoids the oddity of a interface without a public interface
VTable, which may cause problems for language bindings.
Instead of making the frame clock a settable property of a window, make
toplevel windows inherently have a frame clock when created (getting
rid of the default frame clock.) We need to create or destroy frame
clocks when reparenting a window to be a toplevel, or to not be a
toplevel, but otherwise the frame clock for a window is immutable.
Add a very simple GtkWidget function for an "tick" callback, which
is connected to the ::update signal of GdkFrameClock.
Remove:
- GtkTimeline. The consensus is that it is too complex.
- GdkPaintClockTarget. In the rare cases where tick callbacks
aren't sufficient, it's possible to track the
paint clock with ::realize/::unrealize/::hierarchy-changed.
GtkTimeline is kept using ::update directly to allow using a GtkTimeline
with a paint clock but no widget.
If we get a focus event for a X window we don't recognize, just
ignore it and avoid a g-critical when
_gdk_device_manager_core_handle_focus() is called with a NULL window.
Deprecate gdk_window_enable_synchronized_configure() and
gdk_window_configure_done() and make them no-ops. Implement the
handling of _NET_WM_SYNC_REQUEST in terms of the frame cycle -
we know that all processing will be finished in the next frame
cycle after the ConfigureNotify is received.
For an operation like synchronizing audio to video playback, we need to
be able to predict the time that a frame will be presented. The details
of this depend on the windowing system, so make the backend predict
a presentation time for ::begin-frame and set it on the GdkFrameTimings.
The timing algorithm of GdkFrameClockIdle is adjusted to give predictable
presentation times for frames that are not throttled by the windowing
system.
Helper functions:
gdk_frame_clock_get_current_frame_timings()
gdk_frame_clock_get_refresh_info()
are added for operations that would otherwise be needed multiple times
in different locations.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
We want the compositor to do different things for frames where
"slept before" is TRUE. Communicate to the compositor that
frame is a no-delay frame (slept_before=FALSE) by ending the frame
by increasing the counter value by 1, and that the frame is a
normal frame (slept_before=TRUE) by increasing the counter value
by 3.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Add functions that tell us whether the main loop slept before we drew
a frame. Blocking with the frame clock frozen doesn't count as sleeping.
We'll use this to advertise to the compositor whether we
are drawing as fast as possible (and it should do the same) or timing
frames carefully (and it should do the same.)
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Don't start the idle if we're in the middle of painting a frame -
this will prevent us from getting the timing right when starting
the idle after the frame.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
In order to be able to track statistics about how well we are drawing,
and in order to be able to do sophisticated things with frame timing
like predicting per-frame latencies and synchronizing audio with video,
we need to be able to track exactly when previous frames were drawn
to the screen.
Information about each frame is stored in a new GdkFrameTimings object.
A new GdkFrameHistory object is added which keeps a queue of recent
GdkFrameTimings (this is added to avoid further complicating the
implementation of GdkFrameClock.)
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Instead of communicating the start of a frame to the window manager
as soon as we begin a frame, start a frame only when we know we've
actually created damage to the contents of a window.
(This uses cairo_set_mime_data() as a notification mechanism - a
clever suggestion from Uli Schlachter.)
The advantage of this is that we aren't forcing the compositor to
do a frame cycle and send _NET_WM_FRAME_DRAWN - depending on how the
compositor is structured that might either cause it to do extra
work or it might send _NET_WM_FRAME_DRAWN early and upset frame
timing.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Since events can be paused independently for each window during processing,
make _gdk_display_pause_events() count how many times it is called
and only unpause when unpause_events() is called the same number of
times.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Unqueuing events from the windowing system when paused could result
in weird reordering if event filters resulted in application-visible
behavior. Since we now resume events when the frame clock is frozen,
we now no longer count on low-level event handling running while
event handling is paused.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Keeping events paused after the end of a frame put us in a
weird state where we had to process and queue events - so that
we would get the message from the compositor - but not deliver
them. Instead resume events before ending the frame.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
When we have pending motion events, instead of delivering them
directly, request the new FLUSH_EVENTS phase of the frame clock.
This allows us to compress repeated motion events sent to the
same window.
In the FLUSH_EVENTS phase, which occur at priority GDK_PRIORITY_EVENTS + 1,
we deliver any pending motion events then turn off event delivery
until the end of the next frame. Turning off event delivery means
that we'll reliably paint the compressed motion events even if more
have arrived.
Add a motion-compression test case which demonstrates behavior when
an application takes too long handle motion events. It is unusable
without this patch but behaves fine with the patch.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Switch GtkStyleContext to using GdkFrameClock. To do this, add a new
UPDATE phase to GdkFrameClock.
Add a GdkFrameClockTarget interface with a single set_clock() method,
and use this to deal with the fact that GtkWidget only has a frame
clock when realized.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
If the backend is throttling paints, then the frame clock will be
frozen at the end of the frame. If not, then we need to add throttling,
so wait until 16ms after the start of the frame before beginning the
next frame.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
By starting with an odd frame counter value, we make the mapping
and initial paint of the window an atomic operation, avoiding
any visual artifacts from an unpainted window.
Possible improvement: start the frame when doing gdk_window_show(),
so that the same improvement occurs for windows that were previously
shown and are being mapped again.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
When a window is unmapped, freeze its frame clock. This avoids doing
unnecessary work, but also means that we won't block waiting for
_NET_WM_FRAME_DRAWN messages that will never be received since the
frame ended while the window was withdrawn.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
As part of the extended _NET_WM_SYNC_REQUEST_COUNTER protocol,
we get a _NET_WM_FRAME_DRAWN message for each frame we draw. Use this
to synchronize the updates we are doing with the compositing manager's
drawing, and ultimately with with display refresh.
We now set the sync request counters on all windows, including
override-redirect windows, since it is also useful to do synchronized,
atomic updates for such windows.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
By exporting two XSync counters on a toplevel window, we subscribe
to an extended form of the _NET_WM_SYNC_REQUEST_COUNTER protocol,
where the window manager can initiate an atomic frame, as previously,
but the application can also do so by incrementing the new counter to
an odd value, and then to an even value to finish the frame.
See:
https://mail.gnome.org/archives/wm-spec-list/2011-October/msg00006.html
The support for 64-bit integers that GLib requires is used to
simplify the logic.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Add the ability to freeze a frame clock, which pauses its operation,
then thaw it again later to resume.
Initially this is used to implement freezing updates when we are
waiting for ConfigureNotify in response to changing the size of
a toplevel.
We need a per-window clock for this to work properly, so add that
for the X11 backend.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Instead of having gdk_frame_clock_request_frame() have
gdk_frame_clock_request_phase() where we can say what phase we need.
This allows us to know if we get a frame-request during layout whether
it's just a request for drawing from the layout, or whether another
layout phase is needed.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
We may receive events because SubstructureNotifyMask has been selected
for the root window. (Most likely, this would occur because GTK+
is being used inside a window manager like Metacity or Mutter.)
This can confuse various types of internal accounting, so detect
such events and comprehensively ignore them for GDK's internal
purposes. We still need to generate GDK events for these cases
because you can select for substructure events with
GDK_SUBSTRUCTURE_MASK.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
