The current ordering updates the clip shape of the composite overlay
window after unredirecting the target window. This has the effect of
forcing X to clear the target window and sending an expose to the
application to repaint - causing an unsightly flash. If we update the
shape first, then unredirect, X restores the background of the root
window (sending no expose events as no one is interested) and the
background is typically NONE for the root window. Then the unredirect
paints the contents of the composite backing pixmap over top without
requiring a round trip and waiting for the client to repaint - thus no
flashing.
Fixes regression from
commit d6282716b2
Author: Jasper St. Pierre <jstpierre@mecheye.net>
Date: Fri Dec 6 17:10:44 2013 -0500
compositor: Simplify the unredirected window management code
Cc: Jasper St. Pierre <jstpierre@mecheye.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
https://bugzilla.gnome.org/show_bug.cgi?id=743858
This just exposes the type and the singleton getter necessary to make
it available to introspection. We'll expose more functionality as it
becomes needed.
https://bugzilla.gnome.org/show_bug.cgi?id=743745
We're locked to frame sync anyway, so it doesn't make sense to try to
redraw early. In casual testing, this seems to actually make things
faster, as well.
The commit 97a69cee5a broke the caching of
the surface state when because the frame_callback_list target state was
overwritten after the content had been moved to it.
This commit fixes it by moving the frame list addition after the copy. We
also need to initialize the list since the plain copy put garbage in it.
https://bugzilla.gnome.org/show_bug.cgi?id=743678
When looking for space to place a new window, other non-minimized
windows on the same workspace should be taken into account. However
the current check does not work correctly when the placed window is
located on all workspaces, so handle that case explicitly.
https://bugzilla.gnome.org/show_bug.cgi?id=743217
When running as a dispay server pointer barriers are a server side
feature and requires no client interaction of any sort. This patch
implements pointer barriers that can be used when running as a display
server on the native backend. Running as a display server using the X11
backend is currently not supported.
https://bugzilla.gnome.org/show_bug.cgi?id=706655
For each device that can be mapped (touchscreens, tablets), the output
will be fetched from settings and matched with the currently connected
ones. If a match is found, the device matrix will be found out from the
output configuration and set on the device.
This is also updated both individually for newly connected devices, and
collectively on output configuration changes.
https://bugzilla.gnome.org/show_bug.cgi?id=739397
This goes through modifying XI2 device properties, either common ones (eg.
set on every device) or those specific to the libinput X11 driver. Keyboard
repeat/rate are set through core and XKB APIs.
https://bugzilla.gnome.org/show_bug.cgi?id=739397
This object internally keeps track of the relevant input configuration,
and goes through its vmethods in order to apply the configuration on the
backend-specific devices.
So far, only mouse/touchpad settings are actually attached to GSettings
changes. ::set_matrix(), meant for tablets/touchscreens, is not hooked
yet.
One caveat is that meta_input_settings_create() may return NULL if the
backend does not own the windowing system (wayland nested on X11 being
the one case), and thus device settings can't be changed freely.
https://bugzilla.gnome.org/show_bug.cgi?id=739397
This patch removes the X11 specific code from MetaBarrier and creates an
abstraction layer MetaBarrierImpl. The existing X11 implementation is
moved to a new GObject MetaBarrierImplX11 implementing the abstract
interface MetaBarrierImpl which is instantiated by MetaBarrier when
supported.
While at it, move it to backends/ and properly name the files.
https://bugzilla.gnome.org/show_bug.cgi?id=706655
EDID parsing has been refactored to a common meta_output_parse_edid()
function, which ensures the extracted information is the same on both KMS
and X11 backend, so it can be used consistently on eg. settings values.
https://bugzilla.gnome.org/show_bug.cgi?id=742882
MetaKeyCombo is about the *unresolved* keybinding, which can either be a
"keysym" (<Ctrl>F) or a "keycode" (<Ctrl>0x21). When we resolved the
keysym to a keycode, we stuffed it back in the same MetaKeyCombo, which
confused about what the "keycode" field was for. Thus, we often stomped
on the user's explicit choice if they chose a keycode binding value.
To solve this, create a separate structure, the "devirtualized key combo"
or MetaKeyDevirtCombo, which contains a resolved keycode from the
keysym, and a devirtualized modifier mask. The MetaKeyCombo is now
always a "source" value, and the MetaKeyDevirtCombo is now always what
the user chose.
This also lets us significantly clean up the overlay and ISO key binding
paths.
The reason MetaKeyCombo has a keycode value at all is *not* to store the
devirtualized keycode from the keysym, but instead to allow people that
type in "0x55" into the preference. Everything except the overlay-key
respected this. Make the overlay-key binding respect this.
Break down the beautiful core/ui abstraction barrier by inserting
a pointer to MetaWindow into a MetaUIFrame. I'm a scoundrel, I know.
We'll use this very soon to destroy meta_core_get.
Whenever we added a frame to the GHashTable, we added the frame itself
as the value, and a pointer to its storage of the frame window XID,
as the key.
When we iterated over the hash table, we actually looked up the
MetaUIFrame in the key, which might seem extraordinarily wrong, but
eagle-eyed viewers might notice that the XID is the first field in
MetaUIFrame, so the key and value are actually the same pointer.
Changing the layout of MetaUIFrame at all causes this to go haywire,
so let's not do this and simply put the MetaUIFrame in the value,
as expected.
When the frame type updates, we were doing something funky that
caused us to reset the title used for the text layout here. I can't
really think of any place that it would trigger, and in testing I
haven't hit this either, so let's just remove the fancy logic and
assert this.
With support for the old metacity theme format gone, there's no
reason to keep storing theme information in terms of the old theme
properties. Just store the padding/border information for each
element directly.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
MetaFrameStyle now only holds a MetaFrameLayout, so we can cut out
the middle man and use the layout directly. And as we are already
using a single style/layout per frame set and handle frame state
and focus by setting appropriate style flags, MetaFrameStyleSet
is pointless too - just store one MetaFrameLayout per frame type
directly in the theme.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
Rest in peace you magnificent format, love-child of arcane X11 drawing
API and markup craze, you will not be missed.
We do remember however the bravery of a many men and women, who fearlessly
descended into the guts of your intrinsics and turned ugliness into beauty;
their work will still be spoken of when you will long have been forgotten.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
All geometry/drawing information is now picked up from the GTK+ theme,
so replace the remaining bits (hide_buttons + title_scale) with
hardcoded values from the default Adwaita theme and stop loading
the metacity theme altogether.
If there is a need to theme those constants again in the future,
we should make them available from GTK+ where they are available
for client-side decorations as well. They certainly don't justify
maintaining support for a complex theme format.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
Few themes ever had support for those in the first place, and even
less supported them properly; in particular support in the default
theme has been broken for a while now.
With this in mind (and considering that not even the tweak tool exposes
any UI to configure them), let's (try to) remove support altogether - the
corresponding rects are still kept around, so it's easy to add back in
case we reconsider (and get the necessary artwork).
https://bugzilla.gnome.org/show_bug.cgi?id=741917
GTK+ doesn't deal with different frame types for its client-side
decorations - it just treats dialogs the same as normal windows
and ignores the odder frame types like UTILITY and MENU. That's
fine as those have largely gone out of fashion anyway, but it's a
different case for the WM - we still have to support them somehow.
For now, just apply the existing title_scale factor to the geometry
information picked up from the theme in addition to the title font.
If it turns out that there's demand for something more sophisticated,
we can still consider adding wm-only style information to the GTK+
theme.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
The frame shape is relevant in three places:
- the window decoration we draw
- the frame mask (used for the shape region)
- the frame bounds (used for clipping)
All three should match, so make sure to use the same GTK+ method for
the first two, and bring the (non-antialiased) third closer to the
other two by removing an obscure modifier from the corner radius.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
We now have everything in place to pick up geometry and drawing
information from GTK+ rather than the metacity theme, so do just
that; the metacity theme is now only used for some constants
(title_scale, hide_buttons, ...), which we will replace soon.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
We want to eventually pick up all theme information from GTK+ instead
of our own theme format; to prepare for this, add another helper method
to fill in geometry information from the GTK+ theme.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
GTK+ expresses the window state as style classes and widget state for
client-side decorations. Add a helper method to translate our own frame
state to the corresponding changes to the style context hierarchy.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
Sounds obvious, doesn't it?
After this change when titlebar-uses-system-font is set, the "system
font" used will not be a generic one, but match what GTK+ uses in
client-side decorations.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
In order to pick up all theme information from GTK+, a single style
context is not enough; a style hierarchy that closely matches the widget
hierarchy by GTK+'s client-side decorations will allow this soon.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
Our current use of style contexts is fairly limited - we don't
use them for much more than picking up some color information.
We will soon start to make more elaborate use of GTK style
information, but a single context will no longer be enough
to draw a frame then.
To prepare for this, add a simple ref-counted type to wrap
style information.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
Rather than defining the space to the left and right of buttons, add a
simple spacing property that defines the space between buttons, which is
what GTK+ does for client-side decorations (e.g. GtkButtons in a GtkBox).
Unfortunately the value is hardcoded in GTK+; if it is exposed in the
theme in the future, we should pick it up from there, but for now we
just use the same value as GTK+.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
Basically it's odd to have "button_rect" be a function with all the
foo_rect GdkRectangles around - renaming to get_button_rect() will
free the name for the generically named "rect" once buttons are the
only movable pieces in the frame.
https://bugzilla.gnome.org/show_bug.cgi?id=741917
This reverts commit 47e339b46e. The
approach that was used to reduce the amount of work we do on RR events
to the necessary minimum is flawed. It assumes that, when the first
event we see where the retrieved XRRScreenResources.timestamp is
bigger than the previous, we already have all the data we need to
rebuild our view of the world.
That isn't true however, because the X server sends
RRScreenChangeNotify events for every step of the configuration
change, i.e. it lacks an atomic reconfiguration API. In particular, if
the X screen size is one of the changes, when we rebuild our state and
emit monitors-changed, the X screen size might still be the previous
one and since we stop updating ourselves until another reconfiguration
happens (noticed by looking at XRRScreenResources.timestamp) we end up
with the wrong idea of the X screen size.
https://bugzilla.gnome.org/show_bug.cgi?id=738630
This optimization breaks our use of XRRScreenResources' timestamps to
detect hotplugs in case one of the outputs is disconnected and the
remaining ones don't need any mode, position or transform adjustments.
In that scenario, when applying the new configuration, we resize the X
screen but never call XRRSetCrtcConfig() and since XRRSetScreenSize()
doesn't take a timestamp and the X server doesn't update its last set
timestamp, when we next get a RRScreenChangeNotify and update
ourselves, XRRScreenResources.timestamp will still be smaller than
XRRScreenResources.configTimestamp which makes us think we're seeing a
new hotplug. We just don't enter an endless loop because the screen
size that we keep applying is always the same and the X server
short-circuits and stops sending us RRScreenChangeNotifys.
Always calling XRRSetCrtcConfig() ensures that the last set timestamp
will be bigger than configTimestamp in the next event and thus making
us trigger the monitors-changed signal properly.
Note that the X server already does basically the same checks that
we're removing here, so doing this shouldn't be a significant
efficiency loss. See
http://cgit.freedesktop.org/xorg/xserver/tree/randr/rrcrtc.c?h=server-1.16-branch#n539
If the app finished multiple frames before we sent _NET_WM_FRAME_DRAWN,
we could add the send_frame_messages_timer multiple times. In the rare
case that the app immediately closed the window, the older timeout
could potentially then run on the freed actor.
https://bugzilla.gnome.org/show_bug.cgi?id=738686
* Use -1 rather than 0 as a flag for pending queue entries; 0 is
a valid frame_counter value from Cogl.
* Consistently handle the fact we can have more than one pending
entry. It's app misbehavior to submit a new frame before
_NET_WM_FRAME_DRAWN is received; but we accept such frame messages,
so we can't just leak them.
* If we remove send_frame_message_timer, assign the current frame counter
to pending entries.
* To try to avoid regressing on this, when sending _NET_WM_FRAME_TIMINGS
messages, if we have stale messages, or messages with no frame drawn
time, warn and remove them from the queue rather than just accumulating.
* Improve commenting.
https://bugzilla.gnome.org/show_bug.cgi?id=738686
It doesn't make sense to load cursor textures that we might not ever
use. Since the code here also uses CoglTexture2D, and cursors tend
to be NPOT textures, then we won't crash users of cards without
NPOT support. At least until they open the magnifier. :)
Whenever the compositor takes a grab, we're supposed send leave/enter
events to the current surface, which makes sense, as the compositor
has stolen the pointer from the client.
I forget why I added the special case in the first place, but it's
likely a bug that's since been fixed.
This actually fixes a bug: it prevents the need to double-click on
X11 application titlebars when grabbing them.
Windows that set empty input shapes get n_rects of 0 when querying them
later, which makes sense, but the code that interpreted the result
translated it into a NULL input shape, which meant it was the same as
the bounding region. As such, an empty input shape would actually get
interpreted as a full input shape!
We, ourselves, set an empty input shape on tray icon windows in
gnome-shell since we would handle the picking ourselves. This meant that
we'd actually get the MetaSurfaceActorX11 when hovering over the tray
icon, instead of the ShellGTKEmbed that we capture events on and react
to.
This fixes weird tray icon behavior in gnome-shell.
The parent pick() implementation in ClutterActor only recurses if the
vfunc is untouched, which means it's up to the MetaWaylandSurface
implementation to actually recurse, just the same as if an input mask
applied.
https://bugzilla.gnome.org/show_bug.cgi?id=738890
This reverts commit ec8ed1dbb0.
1) It turns out to add a momentary flicker from the transition
between the login screen and user session
2) It actually isn't needed anymore since bug 733026
https://bugzilla.gnome.org/show_bug.cgi?id=740377
Refactor make_default_config() to always sanity-check the configuration to
ensure that it fits within the framebuffer. Previously, this was only done
for the default linear configuration.
In recent versions of the QXL driver, it may set "suggested X|Y" connector
properties. These properties are used to indicate the position at which
multiple displays should be aligned. If all outputs have a suggested position,
the displays are arranged according to these positions, otherwise we fall back
to the default configuration.
At the moment, we trust that the driver has chosen sane values for the
suggested position.
When the output device has hotplug_mode_update (e.g. the qxl driver used in
vms), the displays can be dynamically resized, so the current display
configuration does not often match a stored configuration. When a new
monitor is added, make_default_config() tries to create a new display
configuration by choosing a stored configuration with N-1 monitors, and then
adding a new monitor to the end of the layout. Because the stored config
doesn't match the current outputs, apply_configuration() will routinely
fail, leaving the additional display unconfigured. In this case, it's more
useful to just fall back to creating a new default configuration from
scratch so that all outputs get configured to their preferred mode.
Move logic for creating different types of configurations into separate
functions. This keeps things a bit cleaner and allows us to add alternate
configuration types more easily.
WindowActors can outlive their corresponding window to animate unmap.
Unredirecting the actor does not make sense in that case, so make
sure to not request it.
https://bugzilla.gnome.org/show_bug.cgi?id=740133
When a laptop's lid is closed we try to build and apply a temporary
configuration that disables the laptop's display if we have other
outputs.
This isn't enough though, we must also check if at least one of these
other outputs is enabled otherwise we'll try to resize the screen to
0x0 which (rightfully) hits an assertion.
https://bugzilla.gnome.org/show_bug.cgi?id=739450
The input region currently only gets scaled by the surface
scale while ignoring the output scale, which causes input events to not get
delivered correctly for clients on hidpi screens. So take the output scale
into account when doing so.
https://bugzilla.gnome.org/show_bug.cgi?id=739161
This commit is wrong, it assumes that the scale only applies to the one
set by the client but its not. meta_surface_actor_wayland_scale_texture
also handles the output scale. Revert the commit to fix hidpi for wayland
clients like weston-terminal.
This reverts commit 0364ea9140.
https://bugzilla.gnome.org/show_bug.cgi?id=739161
Since GTK+ commit 3a337156d11a86c7, save()/restore() may only be
used for subelements; in this particular case, the change broke
the backdrop state in decorations. Luckily we don't actually need
the save()/restore() pair anyway, as we only touch the context's
state and always set it explicitly.
The set/unset branches of meta_display_update_pointer_emulating_sequence()
have been split and put directly where it makes sense. The pointer emulated
sequence will be updated before processing the CLUTTER_TOUCH_BEGIN, and
after processing the CLUTTER_TOUCH_END, this way the checks on this hold
true during all the sequence lifetime.
https://bugzilla.gnome.org/show_bug.cgi?id=738411
If the actor surface has an input mask, custom picking is implemented
for the portions affected by the mask, although the child actors (most
usually subsurfaces) are left out.
https://bugzilla.gnome.org/show_bug.cgi?id=738890
Returning FALSE here gets the button release event propagated to the
client on wayland, which is unexpected after xdg_surface.move/resize()
have been called.
https://bugzilla.gnome.org/show_bug.cgi?id=738888
It turns out that this was wrong because MetaWindow->monitor points to
the old monitor infos and they are needed to position windows in the
new configuration which happens in a monitors-changed handler.
This reverts commit e1704acda4.
The code in MetaMonitorConfig was really complex and was trying to do
way too much, using multiple different variables to determine where
things were stored, and trying to do fancy tricks to transfer
ownership.
Add a refcounting system to help simplify this, and clean up the logic.
Simply along the way, this fixes multiple bugs in the monitor config
logic, most notably bug #734889, which was my original goal to fix.
The X server sends several RRScreenChangeNotify events in a burst when
something happens which, currently, causes us to rebuild our view of
the world as many times and notify the upper layers about it which
causes a lot of bogus repeated work like rebuilding background actors.
We can avoid this extra work by looking at the timestamp in the
XRRScreenResources struct which is updated when an X client (including
us!) last changed something and comparing it with the previous
timestamp.
https://bugzilla.gnome.org/show_bug.cgi?id=738630
meta_monitor_config_match_current() only matches the number of outputs
and if the output connector, vendor, product and serial match.
In the X backend, this means that we can't use it to bypass doing any
work because it won't detect cases where we actually want to update
ourselves like e.g. an output being turned off either by us or by
another X client (e.g. xrandr).
In the native backend, unlike the xrandr backend, we only get called
on real hotplug events and thus should always trigger the common
hotplug code to (possibly) apply a new mode so the check is pointless
anyway.
https://bugzilla.gnome.org/show_bug.cgi?id=738630
In randr events, configTimestamp can be considered the hotplug time,
i.e. whenever the server notices hardware changes, this value will be
updated.
Having that in mind, we can re-work the logic to make it clearer.
There are no semantic changes.
Commit 2f9c601 accidentally changed the logic here, changing the grab
behavior when not using raise-on-click. Fix this.
Spotted-by: Adam Goode <adam@spicenitz.org>
From a quick code search and grep of gnome-themes-standard, none of
the themes that I inspected used this feature. Since it's the last
thing that uses a lot of old legacy GdkPixbuf code, I'd rather just
consider the feature unsupported at this point and clean up everything
I need to.
https://bugzilla.gnome.org/show_bug.cgi?id=662962
Fullscreen windows look weird when they are overlapped by system chrome,
which currently happens when another window is stacked above. We used to
auto-minimize fullscreen windows in that case, which proved to be both
unreliable and unpopular. So instead, keep the system chrome hidden even
when the fullscreen window is not stacked at the top.
https://bugzilla.gnome.org/show_bug.cgi?id=693991
We clip the input region to the client rect, so the client rect should
be up to date before we fetch the input region.
This fixes popup windows not working in GTK+2 under Wayland.
We should also update the shape / input regions when the window is
reconfigured for a complete fix, so that making an O-R window bigger
doesn't confuse mutter, but let's leave that to a future commit.
The constructor would collect windows that are sticky before initializing its state
which would lead to a crash in the case of windows with struts which trigger a work
area recalculation where mutter would assume, due to uninitialized state, that an
existing work area has to be freed.
https://bugzilla.gnome.org/show_bug.cgi?id=738384
Damage coordinates are relative to the drawable not to the screen. So we
have to check whether x and y are 0 and not window_rect.x/y otherwise the
herustic will never trigger for windows on monitors whos x and y are not 0.
https://bugzilla.gnome.org/show_bug.cgi?id=738271
The key event should be interpreted by clients with the modifier state
as it was before the event itself just as in X11 input events.
Achieving this in wayland is a matter of sending the key event first
and the modifiers after (if needed).
This isn't really specified in the wayland protocol but it matches
weston's behavior and should avoid corner cases in clients.
https://bugzilla.gnome.org/show_bug.cgi?id=738238
This reverts commit 33acb5fea0.
The issue here is that the pointer actor does not actually get reset
when the actor's reactivity changes, so we end up with stale picks after
actors are destroyed.
I have a local patch to Clutter for this, but I don't have time to
submit it upstream, so let's just use the ugly code for now.
This reverts commit e496ed50d6.
This was incorrect. wl_surface_destructor actually does the full repick
-- doing it here is dangerous, because the destroy listeners actually
run *before* the destructor, not after, so the surface is still alive.
We never want to send pressed keys to wayland clients on enter. The
protocol says that we should send them, presumably so that clients can
trigger their own key repeat routine in case they are given focus and
a key is physically pressed.
Unfortunately this causes some clients, in particular Xwayland, to
register key events that they really shouldn't handle, e.g. on an
Alt+Tab keybinding, where Alt is released before Tab, clients would
see Tab being pressed on enter followed by a key release event for
Tab, meaning that Tab would be processed by the client when it really
shouldn't.
Since the use case for the pressed keys array on enter seems weak to
us, we'll just fake that there are no pressed keys instead which
should be spec compliant even if it might not be true.
https://bugzilla.gnome.org/show_bug.cgi?id=727178
It only contained a pointer to a wl_resource, which isn't much of
value. Just replace it with the wl_resource instead. Any future private
data should be handled by our future role system.
There's a small window before a window that is being unmanaged is
unregistered with the display. The MetaScreen::window-left-monitor
and MetaWorkspace::window-removed emissions fall right into that
window, so code that runs in that time may well be out of our
control; we can make sure that the method it can use to get an
updated list of windows no longer contains the destroyed window
though, which is a much better option than expecting everyone to
filter the list themselves.
The actor is updated on DnD grab motion events, properly notified
when dragging finishes, and destroyed if the client/surface disappear
below its feet.
This actor is a subclass of MetaFeedbackActor that additionally
implements the "drag failed" animation, snapping back to the drag
origin position in a surface.
This actor is a non-reactive container that autoembeds itself into
the feedback window group in the compositor. The API is meant to
help on creating things attached to pointer/touchpoints, with an
X/Y attachment offset, and following the position of certain events.
Although not strictly a window group... This ClutterActor is
meant to stay always on top, and only show non-reactive actors
created by Mutter itself. Two possible usecases for this layer
are DnD surfaces, and touch spots.
We might also want to move cursors out of an overlay in MetaStage
into here at some point.
Keeping track of the surface will be necessary in case it is destroyed
during DnD, and the coordinates will be useful when figuring out the
snap back coordinates.
In Wayland, popup window types are not override-redirect, and thus can
steal window focus away from their parent window when clicked on.
This means that we need to make sure their appearance is properly
propagated to the parent windows so the parent windows don't lose their
focus while they're propagated.
When grabbing with DND, we need to leave the pointer alone and
under the client's control. The code here was a bit messy before about
when it unset the window cursor -- it did it whenever there was no
current surface after repicking, which is a bit wrong, since it will
fire during a drag grab.
Move the check for this to update_cursor_surface, which is our standard
"sync" API for this, and then call update_cursor_surface after we set
the focus.
During a DND grab, pointer->focus_surface is NULL, since the wl_pointer
doesn't have any focused surface (it's in drag mode). In this case, the
drag interface has control of the focus, and when dragging into a NULL
surface, drag_grab_focus won't get called, properly detaching it from
the previous surface.
Let the interface->focus implementation do the fizzling out.
In the future, we should split out wl_pointer's implementation
(pointer->focus_surface) from the Wayland side of the generic pointer
wrapper (pointer->current) and use our event routing system to determine
or similar whether it should go to wl_pointer or wl_data_device.
The code here was a bit messy with the addition of
hotplug_mode_update, and the comments were a bit confusing and
inaccurate. Clean it up and comment it a bit better to make the flow and
intention more clear.
We need to tell clutter's evdev backend about the desktop's key repeat
settings so that our own key bindings event processing and
gnome-shell's chrome widgets get their fake key events for continuous
key press as they expect.
Note that the wayland frontend filters out these events and thus
wayland clients do not see them as specced.
https://bugzilla.gnome.org/show_bug.cgi?id=728055
The X server applies a default keymap to hotplugged keyboard
devices. To enforce our current settings we must re-upload the keymap
when a new keyboard shows up.
Note that setting the VCK keymap causes the server to propagate it
to all slave keyboard devices.
https://bugzilla.gnome.org/show_bug.cgi?id=737673
Since we now directly expose the reverses bindings directly, we
don't have to have this special-case in do_choose_window.
More importantly, if the backwards binding is pressed and has the Shift
key included, this will actually revert it
This doesn't matter for Alt-Tab in gnome-shell, which already replaces
it with a better Alt-Tab replacement, but it does matter for Alt-Esc,
which switches between windows directly.
Windows are relocated before their workspace is removed, however this
is only necessary for windows that are *only* on that workspace; for
windows on all workspaces, that step is annoying as it will unset the
sticky state requested by the user.
https://bugzilla.gnome.org/show_bug.cgi?id=737625
The workspace MRU lists are updated when windows are managed/unmanaged
or change workspaces. However those updates obviously only apply to
existing workspaces - new workspaces will always start out with an empty
MRU list, despite sticky windows already being "on" that workspace.
As we now assert that the list contains all windows located on the
workspace, we need to initialize it correctly to avoid a crash.
https://bugzilla.gnome.org/show_bug.cgi?id=737581
We only grant requests to set the demands-attention hint if the window
is at least partially obscured; so for non-minimized windows on the
active workspace, we check if any other window on the same workspace
that is higher in the stack overlaps.
However in the case of a sticky window, window->workspace is NULL, so
we end up considering any non-sticky window on a different workspace.
At this point we have already established that the window is showing
on the active workspace, so use that to filter for windows that may
overlap.
Since the introduction of set_workspace_state(), window->workspace
will always be NULL when on_all_workspaces is set - passing that
to a workspace function that does not validate its input will then
result in a crash.
Use the get_workspace() function instead, which will always return
a valid workspace.
Since commit 2eec11b445, windows without a __NET_WM_DESKTOP property
that should be on all workspaces are not added to the active workspace;
this is correct, however not adding them to any workspace is not ...
We overrode the property for PowerSaveMode, which meant that gdbus's
auto-generated PropertiesChanged code wasn't being run.
This really confused gnome-rr and gnome-settings-daemon's power plugin
about the current DPMS state of the display, since they used their
cached PowerSaveMode properties, and never saw a PropertiesChanged being
emitted.
If a display was on, they set it to off, and then set it back on, the
setting back on would never fire, since they thought the display was
already off.
To fix this, remove our custom property override and just respond to
notifications on the object.
Namely, this fixes the DPMS management when receiving notifications so
that it now properly times out.
set_workspace_state () returns early when the desired sticky state
and workspace match the current property values, assuming that the
corresponding MRU lists are already correct in that case.
However that might not be the case when we are setting the initial
state, so don't take the shortcut in that case.
https://bugzilla.gnome.org/show_bug.cgi?id=737178
A window may either be sticky because it has been requested as such,
or because it is placed on a non-primary monitor (and the corresponding
preference is set). While we do take the latter into account, we
currently override the sticky state later during initialization;
be a bit more careful there to get the initial state right.
https://bugzilla.gnome.org/show_bug.cgi?id=737178
Use the new DRM capabilities to figure out the correct cursor size, and
make sure that matches instead of hardcoding 64x64. This fixes incorrect
rendering on some newer AMD cards that support 256x256 cursors.
Based heavily on a patch by:
Alvaro Fernando García <alvarofernandogarcia@gmail.com>
Commit 1af0033368 made a subtle change
regarding how XKeysymToKeycode behaves. It does a depth first search
while XKeysymToKeycode is documented to do a breadth first search:
"this function looks in each column of the core keyboard mapping in
turn and returns the lowest numbered key that matches in the lowest
numbered group" - from the XKB library documentation
Looping over all keycodes for each layout and level index makes us go
back to the previous behavior.
https://bugzilla.gnome.org/show_bug.cgi?id=737134
With the change to how hidden windows are stacked, the position
of the guard window with respect to the hidden windows got flipped
and the guard window was at the bottom of everything; fix it to
be on top of the hidden windows.
https://bugzilla.gnome.org/show_bug.cgi?id=737233
Add commands to request the client to minimize or unminimize the window;
unminimize doesn't currently work for GTK+ because it expects XMapRequest
to be received by the window manager, but the window is already mapped.
https://bugzilla.gnome.org/show_bug.cgi?id=737233
Allow putting '|' into the list of windows for assert_stacking to
represent the position of the guard window. Not present is the same
as at the beginning (bottom) of the list.
https://bugzilla.gnome.org/show_bug.cgi?id=737233
Sometimes (for example with minimization) a request from the client
causes queued work rather than immediate work; so make the test client
'wait' command wait for a full frame cycle.
https://bugzilla.gnome.org/show_bug.cgi?id=737233
Commit 2f229c3928 removed the code to compute the above-tab
keycode and replaced it with a simple constant from linux/input.h.
We obviously cannot depend on linux headers on non-linux systems,
so provide a fallback definition in that case (which is expected
to work assuming the system is using the Xorg xf86-input-keyboard
driver).
https://bugzilla.gnome.org/show_bug.cgi?id=737135
The titlebar rect is in window coordinates, while screen regions are
obviously not. Fix by translating into screen coordinates before
testing for overlaps.
https://bugzilla.gnome.org/show_bug.cgi?id=736915
The coordinates in ConfigureNotify *should* be the coordinates of the
client window; using the coordinates of the frame window compensated for
a problem with the interpretation of StaticGravity for some clients but
broke other clients.
This reverts commit f4f70afe31.
https://bugzilla.gnome.org/show_bug.cgi?id=736719
Some applications, like totem, create keyboard/pointer objects from the
same client, and expect it to work. We made this work a while ago, but
due to an oversight in the code, we increment the serial on button press
for every resource that we need to send events to.
Since operations like move/resize use the grab serial of the devices to
determine whether the operation is exact, we need to make sure the same
serial goes to all devices.
Restructure the code so that all that's in the resource loop is the
sending of the event -- all the calculation that's needed happens
outside.
This fixes moving / resizing the Totem window not working sometimes.
https://bugzilla.gnome.org/show_bug.cgi?id=736840
The fix in d61dde1 regressed the position of popup windows, since the
size was 0x0 when we wanted to do a sole move. Only fizzle out in the
path where we actually *do* resize.
https://bugzilla.gnome.org/show_bug.cgi?id=736812
Going through GSD's settings was done in context of patches that
did not land; it is simpler and more consistent with GTK+ to use
the corresponding XSetting instead.
We only broadcast input to the focus_resource_list, so we need to make
sure it's put in the proper list on startup.
This fixes input not working for windows when they first appear.
Argh. There's always more stuff to fix with keyboard/pointer. Every
single time I think I've fixed it, more stuff pops up.
GTK+ requests get_xdg_surface before attaching a buffer, and since it
might take a long time for GTK+ to get around to attaching a buffer and
committing it, our idle for MOVE_RESIZE will kick in beforehand.
And our idle will try to resize the 0x0 window that currently exists,
constrain it to 1x1, which will send a configure event of 1x1 to the
window while it boots up, causing it to awkwardly resize to the minimum
size of the window.
Make sure that in this case, our idle doesn't cause any problems, and
that we fizzle out any idles like this.
The "proper" way to do this would be to delay the creation of the
MetaWindow until a surface is committed, but that's difficult for a
variety of reasons, and might cause unintended issues with focus.
This prevents issues from happening when processing Enter/Leave events
while in another kind of grab op like a Wayland popup or resizing a
window.
This can't ever really happen except outside of a race condition,
with the X server, since we won't ever pass input events to the
X server in any of these cases, but it can't hurt to be more correct
about what the intended operation is.
GTK+ focuses its own windows with RevertToParent, which means that when
a GTK+ CSD window is destroyed, the X server will set the focus back to
the root window. The event stream that we is an UnmapNotify followed by
a FocusOut event. Our own UnmapNotify-handling code unmanages the window
and forcibly changes the focus to the next default window in the stack.
Since UnmapNotify events don't come with timestamps, we query for one,
and set the window focus using that.
But there's *still* a FocusOut event in the stack, with an older
timestamp and serial than our own focusing. We see this, throw it out
since it's older than the most recent focus, but then our own code that
notices the root has been focused kicks in and tries to focus the
default window... using a timestamp older than our most recent focusing.
meta_display_sanity_check_timestamps notices this, and (rightly so)
puts a warning in our face, telling something is awry.
Only let our workarounds kick in when the event is new enough, otherwise
our code will get confused over old events.
This stops the:
Window manager warning: last_focus_time (367917173) is greater than comparison timestamp (367917170). This most likely represents a buggy client sending inaccurate timestamps in messages such as _NET_ACTIVE_WINDOW. Trying to work around...
warning spam when closing a CSD window.
The last_sent size is effectively what size we should send in configure
requests where the size hasn't changed. Thus, if an app commits a new
size, we need to make sure we respect it and don't reconfigure it with
a size it wasn't expecting when the state changes.
This fixes apps being sent a configure event with 0, 0 on startup,
which was confusing Clutter into displaying a 0x0 viewport.
Restacking the frame for a window while unmanaging the window is
harmless, but for undecorated (in particular, client-side-decorated)
windows, this causes problems because the window is typically
destroyed by the client immediately after withredrawing the window.
Skip windows flagged as being unmanaged when assembling the new
stack and when comparing the old order to the new stack.
Add a stacking test for this.
Windows can be freed at some point after they are unmanaged - because
there is an effect in progress, because a language binding is holding
a reference. Therefore, we need to clean up the later to associate
the xwayland and wayland windows deterministically in an "unamanaged"
handler.
https://bugzilla.gnome.org/show_bug.cgi?id=736694
g_idle_add() makes no guarantee about when it will be run - if Mutter
is busy drawing and blocking glXSwapBuffers() it could happen only
minutes later. Use meta_later_add (META_LATER_BEFORE_REDRAW) instead -
this will deterministically be run after the Wayland socket is read
from but before the next frame is painted.
https://bugzilla.gnome.org/show_bug.cgi?id=736694
We shouldn't scale the cursor size in mutter we g-s-d exports
the correct size on hidpi so use gtk-cursor-theme-size.
This way we also catch changes on resolution updates.
https://bugzilla.gnome.org/show_bug.cgi?id=729337
This reverts commit 4fe66ce0a9.
This is wrong ... we should not scale the cursor size but read
the cursor xsettings that gets exported by gsd. Also this won't update on
resolution changes.
https://bugzilla.gnome.org/show_bug.cgi?id=729337
When restacking the last window alone, we would trigger this off-by-one
error. This would throw us off the end of the array, causing lower_below
warnings for nonsensical values.
Since the last window already is lowered below everything else, we
shouldn't need to lower it.
The merge of the commit af46ef3b 'meta_window_new: clean up error handling'
to the wayland branch accidentally added an extra call to meta_error_trap_push(),
meaning that we leaked one level of error traps for each new window.
Fixes warning:
Gdk-WARNING **: XSetErrorHandler() called with a GDK error trap pushed.
https://bugzilla.gnome.org/show_bug.cgi?id=736589
When the screen resizes, we get a configure event for the composite overlay
window - don't pass that to MetaStackTracker, since the COW isn't in the
stack.
Fixes warning:
mutter-WARNING **: STACK_OP_RAISE_ABOVE: window 0x65 not in stack
We have a quite accurate view of the X stack, so there's no good reason to ask
the X server to do restacking that has no effect. (Restackings that have no
effect on either X windows or Wayland windows were generally optimized out in
the synchronization code, but in other cases like moving an X window that is
only beneath Wayland windows to the top of the stack we would make such
requests.)
Removing such requests:
- Is a small efficiency win in itself
- Allows us to immediately go ahead and apply Wayland changes to the verified stack
- Prevents queued Wayland changes piling up waiting for an X event that will never
be received, since the X server will not send confirmation of no-op restacks.
Since such operations may still have an effect on the relative stacking of X
and Wayland windows, we need to continue applying them to the local stack.
https://bugzilla.gnome.org/show_bug.cgi?id=736559
Now that all actual stack shuffle is handled inside stack-tracker.c, we can make
meta_stack_tracker_record_[raise_above/lower_below] internal to that file and
remove the unused meta_stack_tracker_record_lower().
https://bugzilla.gnome.org/show_bug.cgi?id=736559
stack.c:sync_stack_to_xserver had both code for assembling the desired stack, and
code for enforcing the desired stack on the actual stack of X and Wayland windows;
the latter part is properly the domain of stack-tracker.c; moving the code to
apply the stack there both simplifies it and keeps stack.c more manageable.
https://bugzilla.gnome.org/show_bug.cgi?id=736559
There was still code in stack.c to handle skipping override-redirect windows,
but since quite a while ago, meta_stack_add() is not called for OR windows
since they are outside our stacking control. Add an assertion and remove
unnecessary code.
https://bugzilla.gnome.org/show_bug.cgi?id=736559
stack.c kept it's own record of the last stacking it requested, so that
restacking could be done with minimal moves, but we already have a better
view of the stacking order with the stack tracker, so use that instead.
This allows eliminating the special case for the first restack.
https://bugzilla.gnome.org/show_bug.cgi?id=736559
Since MetaStackTracker is the code that knows about the current X stacking order
and the relationship between X windows and Wayland windows, it's cleaner to
encapsulate stack manipulation in MetaStackTracker rather than have the calling
code make the X calls and only call into MetaStackTracker to inform it about
the changes.
https://bugzilla.gnome.org/show_bug.cgi?id=736559
The step where we requery the stacking order from the server than combine
it in an arbitrary fashion with Wayland windows can be eliminated by observing
that we are the final authority for Wayland window stacking - so if we
apply each X event that we receive from the X server to our stack in a
way that leaves the X windows ordered in the same way as on the server,
and apply events that we have stored locally in a way that doesn't affect
the ordering of X windows, than we have a fully correct ordering of windows.
Ordering this in the order of first applying the X event and then applying the
local portion also means that as long as we had an up-to-date view of the X
stack the composite operation will be identical to what was requested.
https://bugzilla.gnome.org/show_bug.cgi?id=736559
Putting X windows and pointers to MetaWindows into a union had a number of
problems:
- It caused awkward initialization and conditionalization
- There was no way to refer to Wayland windows (represented by
MetaWindow *) in the past, which is necessary for the MetaStackTracker
algorithms
- We never even cleaned up old MetaStackWindow so there could be
records in MetaStackWindow pointing to freed MetaWindow.
Replace MetaStackWindow with a 64-bit "stack ID" which is:
- The XID for X Windows
- a "window stamp" for Wayland windows - window stamps are assigned
for all MetaWindow and are unique across the life of the process.
https://bugzilla.gnome.org/show_bug.cgi?id=736559
Add a basic framework for tests of Mutter handling of client behavior;
mutter-test-runner is a Mutter-based compositor that forks off instances
of mutter-test-client and sends commands to them based on scripts.
The scripts also include assertions.
mutter-test-runner always runs in nested-Wayland mode since the separate
copy of Xwayland is helpful in giving a reliably clean X server to
test against.
Initially the commands and assertions are designed to test the stacking
behavior of Mutter, but the framework should be extensible to test other
parts of client behavior like focus.
The tests are installed according to:
https://wiki.gnome.org/Initiatives/GnomeGoals/InstalledTests
if --enable-installed-tests is passed to configure. You can run them
uninstalled with:
cd src && make run-tests
(Not in 'make check' to avoid breaking 'make distcheck' if Mutter can't be
run nested.)
https://bugzilla.gnome.org/show_bug.cgi?id=736505
For reasons related to interaction between the GTK+ CSS code and the
frame sync protocol, the dummy GtkWindow that MetaUI creates to track
theme properties has to be mapped and have MetaWindow associated with it.
Add a private function so that the test framework can filter this out.
https://bugzilla.gnome.org/show_bug.cgi?id=736505
Add a private hook for the test framework to get XSyncAlarmEvent events -
this will be used to implement XSyncCounter based synchronization
so that the test framework can deterministically wait until Mutter
has seen actions performed by an X11 client.
https://bugzilla.gnome.org/show_bug.cgi?id=736505
Add private functions for the test framework to use to find out the
wayland and x11 display names, so they can set up the environment for
children.
https://bugzilla.gnome.org/show_bug.cgi?id=736505
We'll need this in the wayland frontend to send a modifiers event to
clients.
Note that on X11 this isn't needed because key events include the
group index encoded in modifier state. If we ever want to make the
wayland frontend work with the X11 backend we'll handle it then.
https://bugzilla.gnome.org/show_bug.cgi?id=736433
According to the documentation, the method returns "whether the X window
that the actor was displaying has been destroyed" - that is very much
true when we delay the actual actor destruction for a destroy animation,
so update the method accordingly.
https://bugzilla.gnome.org/show_bug.cgi?id=735927
When a window is destroyed, the corresponding actor may still be
kept around for the destroy effect. But as the actor is removed
from the compositor's stack list immediately, the compositor will
always stack it above "valid" window actors - this is not what we
want, so only update the compositor's list when the actor is
actually destroyed.
https://bugzilla.gnome.org/show_bug.cgi?id=735927
Setting the scaling factor immediately after calling clutter_init()
avoids creating the stage at one size, then later resizing it to
a different size.
https://bugzilla.gnome.org/show_bug.cgi?id=736279
In the case of a nested Wayland compositor inside an X session,
Clutter is managing the toplevel window size, so don't call
XResizeWindow on it - this will confuse Clutter and get the size
and the hints out of sync on the toplevel window.
https://bugzilla.gnome.org/show_bug.cgi?id=736279
It's possible for a released pointer to have repick / set_focus on it as
part of sync_input_focus. When the pointer is actually re-init'd, it
will memset 0, which can cause corruption as our destroy listener has
already been added.
Released devices should be idempotent, so just make sure method calls on
them don't have any effect.
We can enter weird states where get_default_window is called during
window unmanagement, before the window has been fully removed from
the stack. Make sure these windows are *never* returned from
get_default_window, as focusing them can cause an assertion fail,
or worse.
If we add device 2, then add device 254, then remove device 254, then
the max device ID will be 253. Scan through all the devices again on
removal to calculate a new max device ID.
Rather than have the DBus code control this, move this into
MetaBackend. This also lets us destroy idle monitors when appropriate,
rather than leaking them forever.
Not having a paint volume causes every single paint to turn into
full-stage redraw, since otherwise culling won't properly work.
Since we don't paint outside of our allocation, just use the simple
default implementation, but also return TRUE inside it.
Make the vignette options properties so they can be animated;
modify the function-call API for meta_background_actor_set_vignette()
to correspond more closely to the 3 properties.
https://bugzilla.gnome.org/show_bug.cgi?id=735637
Without GLSL, we didn't apply the vignetting, which not only made the
background uniform in color, it made it much lighter. Adjust for this
and make the average brightness with the vignette effect the same
with or without GLSL.
https://bugzilla.gnome.org/show_bug.cgi?id=735637
The old requirement that multiple MetaBackgroundActor objects be
layered on top of each to produce blended backgrounds resulted in
extremely inefficient drawing since the entire framebuffer had
to be read and written multiple times.
* Replace the MetaBackground ClutterContent with a plain GObject
that serves to hold the background parameters and prerender
textures to be used to draw the background. It handles
colors, gradients, and blended images, but does not handle
vignetting
* Add vignetting to MetaBackgroundActor directly.
* Add MetaBackgroundImage and MetaBackgroundImageCache to allow
multiple MetaBackground objects to share the same images
By removing the usage of ClutterContent, the following optimizations
were easy to add:
Blending is turned off when the actor is fully opaque
Nearest-neighbour filtering is used when drawing 1:1
The GLSL vignette code is slightly improved to use a vertex shader
snippet for computing the texture coordinate => position in actor
mapping.
https://bugzilla.gnome.org/show_bug.cgi?id=735637
The old check for using NEAREST by checking clutter_actor_is_in_clone_paint()
and meta_actor_is_untransformed (actor) doesn't work properly since
clutter_actor_is_in_clone_paint() does not look at ancestors of the
actor; it only applies to a direct clone of the actor. Using
meta_actor_painting_untransformed() allows us to check exactly what we
care about rather than using tricky approximations.
https://bugzilla.gnome.org/show_bug.cgi?id=735632
The painting_untransformed() function in MetaWindowGroup is useful
elsewhere, in particular if we want to check whether we can avoid
bilinear filtering when painting a texture 1:1.
https://bugzilla.gnome.org/show_bug.cgi?id=735632
Clutter events include the layout index codified into modifier_state,
unlike XI2 device events, which means that we need to mask it out so
that we can match successfully.
A lot of applications assume that the window is fully positioned when it
gets the MapNotify, especially simple applications. Make sure that the
window is only mapped through the calc_showing logic.
meta_surface_actor_is_argb32 assumes that lack of stex means that a window is
ARGB32. When we unredirect a window we detach the texture so we end up without
a texture. Given that should_unredirect returns FALSE when a window is argb32,
we know that this window is indeed not ARGB32.
Returing TRUE in that case causes us to flip between redirected and
unredirected on every paint.
So fix that by returning FALSE in that case.
A wl_surface may have a wl_subsurface interface, but no buffers attached
yet, even though the geometry calculation code for surfaces/subsurfaces
assumes everything has already a buffer.
Just skip subsurfaces that don't have a buffer, those can't be set
a geometry yet, and right now it's crashing accessing the texture from
the NULL surface->buffer.
https://bugzilla.gnome.org/show_bug.cgi?id=735452
When the blended region was empty, meaning we didn't have to paint
anything blended -- the case for an app update -- was drawing the
entire window blended, because of a think-o in the complex and
complicated logic.
Fix this so that we don't draw anything for the blended region when
empty.
O-R windows appear in workspace->windows, which aren't relocatable,
so we can't simply check if the workspace is empty after relocating
all normal windows, since those windows remain.
Make sure that the only windows we have are those that are
on_all_workspaces.
region first
If we're going to render the entire texture blended, then don't bother
painting the unblended stuff, since we're just going to draw on top
anyway.
We indeed call this function if we're not an X11 compositor, but in this
case we're simply calling it to say that we have no cursor overlay. Make
sure not to assert fail in this case.
This makes it so that MetaSurfaceActorWayland is effectively just a
wrapper actor around MetaShapedTexture with some extra scaling. I think
the MetaSurfaceActor subclassing was a bad idea -- we really should have
these abstractions in much higher levels in the stack than the
compositor.
It doesn't make sense to update it in the surface actor. It's also
theoretically wrong to update the buffer's texture on surface commit,
too, because it's buffer state, not surface state, but I don't think
there's any place we use a wl_buffer without a wl_surface.
This is bad behavior, and can also cause us to get in an infinite loop
if an OR window is mapped and unmapped in quick succession. This
sequence causes a MapNotify followed by an UnmapNotify, and when
processing the events, we'll call XMapWindow, XUnmapWindow, which will
put another set of MapNotify, UnmapNotify events in our queue, which we
then process by calling XMapWindow, XUnmapWindow, and so it goes
forever, or at least some scheduler uncorks us by making us call
XMapWindow when the window is already mapped, or XUnmapWindow when the
window is already unmapped.
We can stop this madness by simply making sure never to call neither
XMapWindow or XUnmapWindow on OR windows, which is the correct thing to
do anyway.
This was the original intention, but it was thought to be easier
to mark this as a combination of all directions. It turned out to
instead cause subtle bugs since code that blindly checked & DIR_WEST or
similar turned out to get it wrong when it was UNKNOWN, so just make
it an explicit flag.
This fixes the cursor appearing in the wrong place when starting a
keyboard resize.
