This commits refactors cursor handling code and plugs in logic so that
cursor sprites changes appearance as it moves across the screen.
Renderers are adapted to handle the necessary functionality.
The logic for changing the cursor sprite appearance is done outside of
MetaCursorSprite, and actually where depends on what type of cursor it
is. In mutter we now have two types of cursors that may have their
appearance changed:
- Themed cursors (aka root cursors)
- wl_surface cursors
Themed cursors are created by MetaScreen and when created, when
applicable(*), it will extend the cursor via connecting to a signal
which is emitted everytime the cursor is moved. The signal handler will
calculate the expected scale given the monitor it is on and reload the
theme in a correct size when needed.
wl_surface cursors are created when a wl_surface is assigned the
"cursor" role, i.e. when a client calls wl_pointer.set_cursor. A
cursor role object is created which is connected to the cursor object
by the position signal, and will set a correct texture scale given what
monitor the cursor is on and what scale the wl_surface's active buffer
is in. It will also push new buffers to the same to the cursor object
when new ones are committed to the surface.
This commit also makes texture loading lazy, since the renderer doesn't
calculate a rectangle when the cursor position changes.
The native backend is refactored to be triple-buffered; see the comment
in meta-cursor-renderer-native.c for further explanations.
* when we are running as a Wayland compositor
https://bugzilla.gnome.org/show_bug.cgi?id=744932
Use a specialized cursor renderer when running as a nested Wayand
compositor. This new renderer sets an empty X11 cursor and draws the
cursor as part of the stage using the generic cursor renderer drawing
path.
https://bugzilla.gnome.org/show_bug.cgi?id=744932
Make a surface roles into objects with vfuncs for things where there
before was a big switch statement. The declaration and definition
boilerplate is hidden behind C macros.
https://bugzilla.gnome.org/show_bug.cgi?id=744932
If a surface doesn't have a role, the compositor will not know how, if
or when it will be painted. By adding it to the compositor frame
callback list, the compositor will respond to the client that the
surface has been drawn already which might not be true.
Instead, queue the frame callback in a list that is then processed when
the surface gets a role assigned. The compositor may then, given the
role the surface got, queue the frame callback accordingly.
https://bugzilla.gnome.org/show_bug.cgi?id=744932
Being a "XWayland window" should be considered equivalent to a role,
even though it is not part of any protocol anywhere. The commit doesn't
have any functional difference, but just makes it clear that an
wl_surface managed by XWayland have the same type of special casing as
surface roles as defined by the Wayland protocol.
As the semantics are more explicit given the role is defined, a comment
explaining why the semantics need to be how they are was added.
https://bugzilla.gnome.org/show_bug.cgi?id=744932
If a queued event is being processed after the surface is
destroyed, trying to access the window associated with the surface
will lead to a segmentation fault.
This patch avoids the crash by first checking if the surface is not null.
https://bugzilla.gnome.org/show_bug.cgi?id=754715
This fixes build error caused by commit 614d6bd. We can simply remove
the usage of meta-wayland.c functions in non-wayland build because
META_BACKEND_X11_MODE_NESTED is only used in wayland.
https://bugzilla.gnome.org/show_bug.cgi?id=753948
Use a better name, use GNOME conventions for error handling, open code the
client error reporting and send the error to the correct resource.
wl_subcompositor doesn't have a role error yet, so continue use some
other error. The only effect of this is error received in the client will
be a bit confusing, it will still be disconnected.
https://bugzilla.gnome.org/show_bug.cgi?id=754215
Older GCC only allows "for (int i" in explicit c99 mode - there's probably
no reason that we can't enable that, but avoiding the construct for
a fast fix.
If the user Alt-Tabs out of the window, we will be left thinking
the Alt key is still pressed since we don't see a release for it.
Solve this and other related issues for the nested X11 compositor
by selecting for KeymapStateMask which causes a KeymapNotify event
to be sent after each FocusIn, and when we get these events, update
the internal XKB state and send any necessary modifiers events to
clients.
https://bugzilla.gnome.org/show_bug.cgi?id=753948
Some backgrounds don't fully fill the screen. For those backgrounds
it's important to paint a color behind them to fill in the gaps.
This commit checks whether or not the background image textures take
up the entire monitor, and in the event they don't, draws a color
behind them (such as it would do if the background were
translucent).
https://bugzilla.gnome.org/show_bug.cgi?id=754476
meta_background_get_texture only draws the bottom image texture
if
1) the blend factor leaves the top image translucent
or
2) the top image is translucent from alpha
The latter case doesn't actually matter since we're using REPLACE
on the top image texture.
This commit drops the unnecessary check for the second case and
applies demorgans law to the conditional for clarity.
https://bugzilla.gnome.org/show_bug.cgi?id=754476
There were lots of code handling the native renderer specific cases;
move these parts to the renderer. Note that this causes the X11 case to
always generate the texture which is a waste of memory, but his
regression will be fixed in a following commit.
The lazy loading of the texture was removed because it was eventually
always loaded anyway indirectly by the renderer to calculate the
current rect.
https://bugzilla.gnome.org/show_bug.cgi?id=744932
There is nothing special about the private API which only consists of
getters for renderer specific backing buffer. Lets them to the regular
.h file and treat them as part of the normal API.
https://bugzilla.gnome.org/show_bug.cgi?id=744932
When a client binds an incompatible version, we should terminate it.
This check should only be there for the unstable version, as once it is
declared stable and renamed, future versions will be backward compatible.
https://bugzilla.gnome.org/show_bug.cgi?id=753855
Before, it used to be in the screen, but now,
meta_cursor_reference_from_theme can never fail. Move it to where we
load the images from the cursor name.
meta_wayland_pointer_get_client_pointer() may be called when the
MetaWaylandPointer as been already shut down, so the hash table will be
NULL at that moment.
Since mutter has two X connections and does damage handling on the
frontend while fence triggering is done on the backend, we have a race
between XDamageSubtract() and XSyncFenceTrigger() causing missed
redraws in the GL_EXT_X11_sync_object path.
If the fence trigger gets processed first by the server, any client
drawing that happens between that and the damage subtract being
processed and is completely contained in the last damage event box
that mutter got, won't be included in the current frame nor will it
cause a new damage event.
A simple fix for this would be XSync()ing on the frontend connection
after doing all the damage subtracts but that would add a round trip
on every frame again which defeats the asynchronous design of X
fences.
Instead, if we move fence handling to the frontend we automatically
get the right ordering between damage subtracts and fence triggers.
https://bugzilla.gnome.org/show_bug.cgi?id=728464
The global wl_pointer_gestures object is now created, effectively
bridging pinch/swipe gestures with clients, so they're now
accessible to clients implementing the protocol.
Instead of moving around all the bound pointer resources for a client
when changing focus, keep all the resources bound by a client in a per
client struct, and track the focus by having a pointer to the current
active pointer client struct instance.
This will simplify having wl_pointer extensinos sharing the pointer
focus of the wl_pointer by only having to add them to the pointer
client.
https://bugzilla.gnome.org/show_bug.cgi?id=744104
While we shouldn't normally receive crossing events for any windows
except the stage when running nested, we do in case we hold a pointer
grab - just ignore those events instead of taking down the user's
session.
The spec says:
"A server should avoid signalling the frame callbacks if the surface is not
visible in any way, e.g. the surface is off-screen, or completely obscured
by other opaque surfaces."
We actually do have the information to do that but we are always calling
the frame callbacks in after_stage_paint. So fix that to only call when
when the surface gets drawn on screen.
https://bugzilla.gnome.org/show_bug.cgi?id=739163
The compositor maintains a ring of shared fences with the X server in order to
properly synchronize rendering between the X server and the compositor's GPU
channel. When all of the fences have been used, the compositor needs to reset
one so that it can be reused. It does this by first waiting on the CPU for the
fence to become triggered, and then sending a request to the X server to reset
the fence.
If the compositor's GPU channel is busy processing other work (e.g. the desktop
switcher animation), then the X server may process the reset request before the
GPU has consumed the fence. This causes the GPU channel to hang.
Fix the problem by having the compositor's GPU channel trigger its own fence
after waiting for the X server's fence. Wait for that fence on the CPU before
sending the reset request to the X server. This ensures that the GPU has
consumed the X11 fence before the server resets it.
Signed-off-by: Aaron Plattner <aplattner@nvidia.com>
https://bugzilla.gnome.org/show_bug.cgi?id=728464
If GL advertises this extension we'll use it to synchronize X with GL
rendering instead of relying on the XSync() behavior with open source
drivers.
Some driver bugs were uncovered while working on this so if we have
had to reboot the ring a few times, something is probably wrong and
we're likely to just make things worse by continuing to try. Let's
err on the side of caution, disable ourselves and fallback to the
XSync() path in the compositor.
https://bugzilla.gnome.org/show_bug.cgi?id=728464
If we can't put up a popup because grabbing the pointer fails we
immediately dismiss the popup but the client might have made requests
already, in particular it might have commited the surface and in that
case we should ignore it.
https://bugzilla.gnome.org/show_bug.cgi?id=753237
When a client sets an input region or a opaque region to NULL, it
should still be considered a change to the corresponding region on the
actor. This patch makes sure this state is properly forwarded.
https://bugzilla.gnome.org/show_bug.cgi?id=753222
This was introduced in commit c6793d477a
to prevent window self-maximisation. It turns out that that bug seems
to have been fixed meanwhile in a different way since the reproducer
in https://bugzilla.gnome.org/show_bug.cgi?id=461927#c37 now works
fine with this special handling removed.
In fact, failing to set window->fullscreen immediately when loading
the initial set of X properties causes us to create a UI frame for a
window that sets _NET_WM_STATE_FULLSCREEN.
This, in turn, might cause the fullscreen constrain code to fail if
the window also sets min_width/min_height size hints to be the monitor
size since the UI frame size added to those makes the rectangle too
big to fit the monitor. If the window doesn't set these hints, we
fullscreen it but the window will get sized such that the UI frame is
taken into account while it really shouldn't (see the reproducer
above).
https://bugzilla.gnome.org/show_bug.cgi?id=753020
Since commit 14b0a83f64 we store the
main window monitor instead of computing it every time. This means
that we must now ensure that it's updated before trying to use it
which we do from meta_screen_resize_func() or else we'll crash on an
assertion later on when removing a monitor:
assertion failed: (which_monitor < workspace->screen->n_monitor_infos)
https://bugzilla.gnome.org/show_bug.cgi?id=752674
Some monitors return a bunch of bytes on their display descriptor
which aren't valid utf8 and thus we fail to serialize them later on
for the DisplayConfig DBus API.
Let's fall back to the stringified product code and serial number in
that case.
https://bugzilla.gnome.org/show_bug.cgi?id=752673
There's a chance the icon will be animated, so store the XcursorImages
instead of the individual XcursorImage, and handle that as a nimages=1
special case.
API to "tick" a cursor animation, and retrieve current frame timing
information has been added.
https://bugzilla.gnome.org/show_bug.cgi?id=752342
They otherwise fall through paths that enable bypass_clutter, this
is necessary so they can be picked by captured-event handlers
along the actor hierarchy.
https://bugzilla.gnome.org/show_bug.cgi?id=752248
When placing a popup and the legacy transient wl_shell_surface surfaces,
take the current scale of the window into account. This commit doesn't
fix relative positioning in case a window scale would change, but since
the use case for relative positioning is mostly popups, which would be
dismissed before the parent window would be moved, it should not be that
much of a problem.
https://bugzilla.gnome.org/show_bug.cgi?id=744934
Make meta_wayland_surface_get_toplevel_window return the top most window
in case its a chain of popups. This is to make all popups in a chain
including the top most surface have the same scale.
The reason for this is that popups are mostly integrated part of the
user interface of its parent (such as menus). Having them in a different
scale would look awkward.
Note that this doesn't affect non-popup windows with parent-child
relationship, because such windows are typically not an integral part of
the user interface (settings window, dialogs, ..) and can typically be
moved independently. It would probably make sense to make attached modal
dialogs have the same scale as their parent windows, but modal dialogs
are currently not supported for Wayland clients.
https://bugzilla.gnome.org/show_bug.cgi?id=744934
Since we scale surface actors given what main output their toplevel
window is on, also scale the window geometry coordinates and sizes
(window->rect size and window->custom_frame_extents.top/left) in order
to make the window geometry represent what is being rendered on the
stage.
https://bugzilla.gnome.org/show_bug.cgi?id=744934
The main monitor of a window is maintained as 'window->monitor' and is
updated when the window is resized or moved. Lets avoid calculating it
every time it`s needed.
https://bugzilla.gnome.org/show_bug.cgi?id=744934
Tracking back from the monitor to the output every time we need to
figure out the scale of a window on a monitor is inconvenient, so
propagate the scale from the output to the monitor it is associated
with.
https://bugzilla.gnome.org/show_bug.cgi?id=744934
A MetaWaylandSurface was casted into a ClutterActor, but it should have
been the MetaSurfaceActor.
Move out parent_actor and surface_actor out of the loop while at it
since they won't change when iterating.
https://bugzilla.gnome.org/show_bug.cgi?id=745655
Keep the active position state in its original coordinate space, and
synchronize the surface actor with it when it changes and when
synchronizing the rest of the surface state, in case the surface scale
had changed.
https://bugzilla.gnome.org/show_bug.cgi?id=745655
We can get this operation in some cases, for example when
we're trying to resize window that cannot be resized.
This can occur with maximized windows that have a border
(without border we couldn't resize them by mouse in maximized state).
In this case we reached abort() beacuse we did not handle this op.
https://bugzilla.gnome.org/show_bug.cgi?id=751884
Previously a MetaWaylandOutput could be removed from the current outputs
table (by being unplugged for example). This would result in the global
object being removed and the MetaWaylandOutput instance freed, but the
wl_resource destructor would still try to remove itself from the list of
resources. Trying to do this, it'd try to access its user data pointer
which would point to the freed MetaWaylandOutput instance, and as a
result crash when trying to manipulate the freed data.
https://bugzilla.gnome.org/show_bug.cgi?id=744453
Whenever a MetaSurfaceActor is painted, update the list of what outputs
the surface is being drawed upon. Since we do this on paint, we
effectively avoids this whenever the surface is not drawn, for example
being minimized, on a non-active workspace, or simply outside of the
damage region of a frame.
DND icons and cursors are not affected by this patch, since they are not
drawn as MetaSurfaceActors. If a MetaSurfaceActor or a parent is cloned,
then we'll check the position of the original actor again when the clone is
drawn, which is slightly expensive, but harmless. If the MetaShapedTexture
instead is cloned, as GNOME Shell does in many cases, then these clones
will not cause duplicate position checks.
https://bugzilla.gnome.org/show_bug.cgi?id=744453
We may access it during painting even if it has been freed. For now,
manually unset it during the MetaWaylandSurface cleanup; in the future
make MetaWaylandSurface a GObject and make the surface pointer a weak
reference.
https://bugzilla.gnome.org/show_bug.cgi?id=744453
Lets use the role when doing role specific commit actions. The
conditions effectively do that anyway, and this way we will get a
compiler warning here whenever we add a new role, as well as we avoid
having different variants of role-determination checks in different
places.
https://bugzilla.gnome.org/show_bug.cgi?id=744453
Enable a user to test and debug multi output configurations on Wayland
without having the available hardware by enabling some basic
configuration of the dummy monitor manager.
Currently available configuration options are:
MUTTER_DEBUG_NUM_DUMMY_MONITORS - to set the number of monitors
MUTTER_DEBUG_DUMMY_MONITOR_SCALES - to configure the monitor scales
See src/backends/meta-monitor-manager-dummy.c for detailed description
of the available configuration parameters.
https://bugzilla.gnome.org/show_bug.cgi?id=747089
Before submitting a new scroll mode, click method or sendevents mode check if
the value is supported by the device. This avoids BadValue errors when setting
two-finger scrolling on single-finger touchpad devices since we can't easily
handle BadValue (see 9747277b)
https://bugzilla.gnome.org/show_bug.cgi?id=750816
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The elementary guys would like this as an API, and I don't see any
reason to refuse -- this is quite nice shadow painting code :)
For some reason, gobject-introspection can't seem to cope with
MetaWindowShape. I'll look into it a bit later, but for now, mark
the function it has trouble with as (skip).
It seems that when translated, paint_offset and actor_offset will always
be the same, so our translation of the clip group won't work. For now,
until I figure out what's going on here, just use the painting offset,
since that what seems to make sense to me.
I didn't write this code, though, so I don't know why the actor's
allocation was involved in this computation at all.
I tested briefly with clones (magnifier, manual cloning through the
looking glass) and couldn't find any other artifacts, so I'm going to do
this for now.
A much less hacky version of maximize / unmaximize is reimplemented
in terms of this, but it could also eventually be used for fullscreen /
unfullscreen, and tile / untile.
The comment explains it better, but Clutter tries to be smart and
repaint actors when their allocations change. Since the window group's
allocation changes when windows move around, this means that moving a
window will always cause a full-stage repaint, which is super slow.
Hack around this for now.
While nothing will completely fix X11's artifacts, this tends to look a
bit better, *especially* with mask textures that have black at the
edges (which are most of them).
It's also faster for GPUs to manage.
The only time we ever execute this code is when we're minimizing or
hiding a window, in which case we should respect stacking order.
This fixes weird "bugs" where windows from the same app magically pop up
over other windows.
We should not be setting random output properties like this.
Use the function we just introduced to only set the underscan flag when
it's actually supported.
So that clients such as the control center can decide to hide an
underscanning checkbutton when the output does not support it.
Support in the KMS / native backend to come later...
This is an extremely niche feature, and conflicts with the rest of our
interface being consistent about not allowing resizing while tiled or
maximized.
A window may be hidden even if not minimized itself, for instance
when an ancestor is minimized. As meta_window_focus() will refuse
to actually focus the window in that case, don't pick it in the first
place.
https://bugzilla.gnome.org/show_bug.cgi?id=751715
Firstly, this patch makes MetawaylandDataSource a GObject. This is in
order to easier track its lifetime without adding destroy signals etc. It
also makes the vfunc table GObject class functions instead while at it,
as well as moves protocol specific part of the source into their own
implementations.
An important part of this patch is the change of ownership. Prior to this
patch, MetaWaylandDataDevice would kind of own the source, but for
Wayland sources it would remove it if the corresponding wl_resource was
destroyed. For XWayland clients it would own it completely, and only
remove it if the source was replaced.
This patch changes so that the protocol implementation owns the source.
For Wayland sources, the wl_resource owns the source, and the
MetaWaylandDataDevice sets a weak reference (so in other words, no
semantical changes really). For XWayland sources, the source is owned by
the selection bridge, and not removed until replaced or if the client
goes away.
Given the changes in ownership, data offers may now properly track the
lifetime of a source it represents. Prior to this patch, if an offer with
an XWayland source would loose its source, it wouldn't get notified and
have an invalid pointer it would potentally crash on. For Wayland
sources, an offer would have a weak reference and clean itself up if the
source went away. This patch changes so the behavior is consistent,
meaning a weak reference is added to the source GObject so that the offer
can behave correctly both for Wayland sources and XWayland sources.
https://bugzilla.gnome.org/show_bug.cgi?id=750680
When a possible drag dest client crashes during DnD, it may happen
we receive first the destroy notification for the data_device, and
later the notification for the focus surface. When this happens we
unset the drag_focus_data_device first, and later on
meta_wayland_drag_grab_set_focus(grab, NULL) we assume it still
exists when sending the leave event, leading to mutter crashing
right after.
So, as we don't receive any ordering guarantees about resource
destruction, just prepare the meta_wayland_drag_grab_set_focus()
paths for this.
It seems that fglrx sometimes gives us absolute junk when requesting the
outputs, and if we don't trap errors, we'll just crash when trying to
configure a junk output. Use xcb so errors simply get ignored.
For enter / leave events, which we use in the UI code, we need to make
sure that these coordinates are root-relative as well, otherwise the
cursor when entering frames might be incorrect.
Going from fullscreen to unfullscreen involves a frame border size, so
in order to properly interpret the saved rect size, we need to make sure
that the frame borders are fully up to date.
When we're unredirected, we don't have a pixmap, and thus our allocation
becomes 0x0. So when events come in, they pass right through our actor,
going to the one underneath in the stack.
Fix this by having a fallback size on the shaped texture actor when
we're unredirected, causing it to always have a valid allocation.
This fixes clicking on stuff in sloppy / mouse mode focus.
In commit cc5def1, buttons were changed from GdkRectangles to
MetaButtonSpace units, but the corresponding memset hack was not.
This means that the clickable portion of the unshade rectangle
was always set to uninitalized memory. The effects of this were
random, but in cases where the moon is aligned just right, the
rectangle would graze over the borders, and so it would take priority
over other borders and show a pointer cursor instead of a resize
cursor.
Since we're using xcb now, not Xlib, it doesn't have the stupid silly
thing where it copies all 32-bit data to 64-bit data to match the
long-y-ness of it.
It seems the largest possible value is to be passed, so actually pass
that instead. Note that even though the name of the xcb_get_property
parameter is called 'long_length' its actually a uint32_t.
https://bugzilla.gnome.org/show_bug.cgi?id=751136
If a client unsets a selection (calls set_selection with the offer
NULL), this should cause the compositor not to continue sending the
previously set offer.
https://bugzilla.gnome.org/show_bug.cgi?id=750007
If we're running as a nested compositor, we must not attempt to
passive grab on the root window, and we should be setting the
touch event mask on the stage window.
https://bugzilla.gnome.org/show_bug.cgi?id=751036
The "calc showing" operation is queued in a few places alongside MetaWindow
creation, we should be ignoring these until there is a buffer to show.
https://bugzilla.gnome.org/show_bug.cgi?id=750552
Mutter generates a motion event for every button and scroll events,
which confuses Xwayland apps that rely on XMotionEvents for various
purposes, e.g. it fools rxvt jumpy mouse detection code.
Remove the call to notify_motion() from the button and scroll event
handlers to avoid these spurious motion events.
https://bugzilla.gnome.org/show_bug.cgi?id=748705
This way, we won't be hit with BadValue errors if we set it to a value
outside the X device's range. This can happen for touchpads without
two-finger scrolling, for instance.
According to the xdg-shell protocol specification the (x, y) coordinates
passed when creating a popup surface is relative to top left corner of
the parent surface, but prior to this patch, if the parent surface was
a xdg_surface, it'd position it relative to top left corner of the
window geometry of that xdg_surface.
https://bugzilla.gnome.org/show_bug.cgi?id=749716
Instead of selecting the first drm mode as the preferred mode, select the
first drm mode marked as preferred. If there are no modes marked as
preferred, revert to the old behaviour and select the first mode.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
https://bugzilla.gnome.org/show_bug.cgi?id=750363
Read the drm layout properties suggested_X, suggested_Y and
hotplug_mode_update and transfer them to the meta layer.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
https://bugzilla.gnome.org/show_bug.cgi?id=750363
The monitors info structure is created from the tiled outputs
and this is used as the central storage for info about a monitor
as opposed to the output state.
It appears at least the EDID mm w/h is for the whole monitor and
not per tile.
this just adds backend support for retrieving the tile
information from X11 (randr 1.5) and native backends.
It stores the tiling information into the output struct.
