Opening and closing the device may result into XI2 grabs being cut short,
resulting into pad buttons being rendered ineffective, and other possible
misbehaviors. This is an XInput flaw that fell in the gap between XI1 and
XI2, and has no easy fix. It pays us for mixing both versions, I guess...
Work this around by keeping the XI1 XDevice attached to the
ClutterInputDevice, this way it will live long enough that this is not
a concern.
Investigation of this bug was mostly carried by Peter Hutterer, I'm just
the executing hand.
https://gitlab.gnome.org/GNOME/mutter/issues/7Closes: #7
A comparison in translate_device_event() does not account for the fact
that X's clock wraps about every 49.7 days. When triggered, this causes
an unresponsive GUI.
Replace simple less-than comparison with XSERVER_TIME_IS_BEFORE macro,
which accounts for the wrapping of X's clock.
Closes: https://gitlab.gnome.org/GNOME/mutter/issues/12
The tertiary-button-action (see bug 790028) is a place for g-c-c to store
the action which should be performed when a stylus' third button is pressed.
Pressing this button is signaled as a BTN_STYLUS3 event from the kernel or
X11 button 8.
https://bugzilla.gnome.org/show_bug.cgi?id=790033
When a window's workspace is not NULL, on_all_workspace should be FALSE.
Similarly, when on_all_workspace is TRUE, the window workspace should be
NULL.
This is an assumption in multiple places in the code, including when
setting the workspace state, the window is either added or removed from
all workspaces only if the window's workspace is NULL.
This rule is initially enforced at creation in _meta_window_shared_new()
when a initial workspace is set. However, when the initial workspace is
set from the session info, the initial workspace is not marked as “set”
which leads to an assertion failure when unmanaging windows, because the
window is not removed from all the workspaces.
When applying the session info to a window, mark the workspace as “set”.
https://gitlab.gnome.org/GNOME/mutter/issues/4Closes: #4
Having “on_all_workspaces_requested” FALSE on a window does not imply a
workspace is set.
If the X11 window is placed on a secondary monitor while workspaces
applies on primary monitor only (“workspaces-only-on-primary” set) then
“on_all_workspaces_requested” is FALSE while “on_all_workspaces“ is TRUE
and the associated workspace is NULL, leading to a crash when saving the
gnome-shell/mutter session.
So if no workspace is set, use the “initial_workspace” instead to avoid
a NULL pointer dereference.
https://bugzilla.gnome.org/show_bug.cgi?id=792818
Now that we have the list of supported modifiers from the monitor
manager (via the CRTCs to the primary planes), we can use this to inform
EGL it can use those modifiers to allocate the GBM surface with. Doing
so allows us to use tiling and compression for our scanout surfaces.
This requires the Mesa commit in:
Mesa 10.3 (08264e5dad4df448e7718e782ad9077902089a07) or
Mesa 10.2.7 (55d28925e6109a4afd61f109e845a8a51bd17652).
Otherwise Mesa closes the fd behind our back and re-importing will fail.
See FDO bug #76188 for details.
https://bugzilla.gnome.org/show_bug.cgi?id=785779
Newer versions of GBM support buffer modifiers, including multi-plane
buffers. Use this new API to explicitly pull the information from GBM,
and feed it to drmModeAddFB2WithModifiers.
https://bugzilla.gnome.org/show_bug.cgi?id=785779
The KMS IN_FORMATS blob property contains a structure defining which
format/modifier combinations are supported for each plane. Use this to
extract a list of acceptable modifiers to use for the primary plane for
XRGB8888, so we can ask EGL to allocate tiled/compressed buffers for
scanout when available.
https://bugzilla.gnome.org/show_bug.cgi?id=785779
Using 800x600 as minimum logical size is very 4:3 thinking, while a lot of
modern devices are 16:9. The specific reason for this commit is to allow
1.5 scaling at mini-laptops (clamshell devices) with e.g. a 5.5"
1280x720 screen. Given that this device has a keyboard, one obviously
is not holding it very close to ones eyes and at 220 dpi that means the text
is too small at scale 1.0. For one real world example of such a device see:
https://en.wikipedia.org/wiki/GPD_Winhttps://bugzilla.gnome.org/show_bug.cgi?id=792765
Offer the text-input interface global, so it can be used by clients. The
MetaWaylandSeat will also let MetaWaylandTextInput intercept key events
before the keyboard interface handles those.
This is the implementation of the internal text-input protocol that will
be used to communicate IMs (to be implemented by gnome-shell) with clients.
The text_input protocol has its own focus expressed through enter/leave
events, that will typically follow the keyboard's.
The client will be able to communicate its current status (eg. focus state,
cursor rectangle in surface coordinates, text surrounding the cursor
position, ...) and will receive commands from the compositor (eg. preedit
text, committing a string, ...).
Whenever there is an active input method, the compositor will route key
events directly through it. The client will not receive wl_keyboard
events if the event is consumed by the IM.
Issuing a shortcut inhibit request for a surface without a window set
will lead to a crash when trying to show the shortcut inhibitor dialog.
In such a case, it's safer to deny the request.
https://bugzilla.gnome.org/show_bug.cgi?id=792599
The events might fall through if there's no corresponding active
pointer/keyboard/touch interface. Barring bugs this should be safe to do,
just a bit wasteful.
When maximizing a window, the previous location is saved so that
un-maximize would restore the same original window location.
However, if a Wayland client starts with a window maximized, the
previous location will be 0x0, so if we have to force placement in
xdg_toplevel_set_maximized(), we should update the location as well so
that the window is placed on the right monitor when un-maximizing.
For that purpose, add a new flag to force the update of the window
location, and use that flag from xdg_toplevel_set_maximized().
https://bugzilla.gnome.org/show_bug.cgi?id=783901
Wayland clients know their size better, so for Wayland we'd rather not
try to resize the client on un-maximize, but for this to work we need a
new MetaMoveResizeFlags.
https://bugzilla.gnome.org/show_bug.cgi?id=783901
When closing a window and showing a new one, the new one may not be
granted input focus until it gets a buffer on Wayland.
If another window is chosen to receive focus and raised on top of stack,
the newly mapped window is focused but placed underneath that other
window.
Meaning that for Wayland surfaces, we need to defer adding the window to
the stack until we actually get to show it, once we have a buffer
attached.
Rather that checking the windowing backend prior to decide if a window
is stackable or not, introduce a new vfunc is_stackable() which tells
if a window should be added to the stack regardless of the underlying
windowing system.
Also add meta_window_is_in_stack() API rather than checking the stack
position directly (replacing the define WINDOW_IN_STACK only available
in stack.c) and remove a window from the stack only if it is present
in the stack, so that the test in meta_stack_remote() becomes
irrelevant.
https://bugzilla.gnome.org/show_bug.cgi?id=780820
Wayland clients using the wl_shell interface were never receiving mouse
input. It meant they also couldn't be raised with a click.
This was because the call to meta_wayland_surface_set_window for wl_shell
surfaces did nothing while surface->window == window already. As such, it
never called clutter_actor_set_reactive() and the wl_shell window remained
a non-reactive actor.
Just make sure surface->window isn't already set before calling
meta_wayland_surface_set_window so it can actually do what it's meant to.
https://bugzilla.gnome.org/show_bug.cgi?id=790309
The device orientation coming out of iio-sensor-proxy defines upright/normal
as the direction in which the picture is displayed on the LCD panel without
any rotation. This is necessary for accelerometer rotation to work properly
in desktop environments which are not aware of panel-orientation issues.
This means that we need to correct the logical-monitor-config / user-visible
rotation for the panel-orientation when we get rotation info from
iio-sensor-proxy.
https://bugzilla.gnome.org/show_bug.cgi?id=782294
Just like we swap the x and y resolution of the monitor modes when
the panel-orientation requires 90 or 270 degree rotation to compensate,
we should do the same for the width and height in mm of the monitor.
https://bugzilla.gnome.org/show_bug.cgi?id=782294
If a monitor's max resolution is a portrait resolution, then assume it is
a native portrait monitor and add portrait versions of the common modes.
https://bugzilla.gnome.org/show_bug.cgi?id=782294
Even if the logical_monitor config does not have an active transform,
we might still be doing a transform under the hood to compensate for
panel-orientation. Check for this and fall back to the sw cursor if this
is the case.
https://bugzilla.gnome.org/show_bug.cgi?id=782294
If a LCD panel has a non normal orientation (mounted upside-down or 90
degrees rotated) then the kernel will report touchscreen coordinates with
the origin matching the native (e.g. upside down) coordinates of the panel.
Since we transparently rotate the image on the panel to correct for the
non normal panel-orientation, we must apply the same transform to input
coordinates to keep the aligned.
https://bugzilla.gnome.org/show_bug.cgi?id=782294
Some x86 clamshell design devices use portrait tablet LCD panels while
they should use a landscape panel, resoluting in a 90 degree rotated
picture.
Newer kernels detect this and rotate the fb console in software to
compensate. These kernels also export their knowledge of the LCD panel
orientation vs the casing in a "panel orientation" drm_connector property.
This commit adds support to mutter for reading the "panel orientation"
and transparently (from a mutter consumer's pov) fixing this by applying
a (hidden) rotation transform to compensate for the panel orientation.
Related: https://bugs.freedesktop.org/show_bug.cgi?id=94894https://bugzilla.gnome.org/show_bug.cgi?id=782294
If input happens to be grabbed somewhere along the shell, and ungrabbed
while a touch operation is ongoing, the wayland bits will happily start
sending wl_touch.update events from an undeterminate point, without
clients having ever received wl_touch.down for that id.
Consider those touches grabbed for the entirety of their lifetime, if
wl_touch.down wasn't received by the client, no other events will.
https://bugzilla.gnome.org/show_bug.cgi?id=776220
When a Wayland client issues a shortcut inhibit request which is granted
by the user, the Super key should be passed to the surface instead of
being handled by the compositor.
https://bugzilla.gnome.org/show_bug.cgi?id=790627
On VT switch, the xkb state layout index is lost and reset to the first
group, so if the first layout is not the last one being used, the xkb
state used in both meta-wayland-keyboard.c and clutter/evdev will be
desynchronized with the keyboard source indicator in the gnome-shell UI.
Save the effective layout chosen along with the seat so it can be
restored when reclaiming devices.
Use the saved layout index from the clutter/evdev's seat to restore the
layout in meta-wayland-keyboard, so that switching VT doesn't reset the
layout and causes further discrepancies with the layout indicator in the
gnome-shell UI.
https://bugzilla.gnome.org/show_bug.cgi?id=791383
The reason why multiple keycodes could be mapped to a single keysym was
to support having both KEY_FAVORITES and KEY_BOOKMARK map to
XF86Favorites. However, iterating through all layout levels adding all
key codes has severe consequences on layouts with levels that map
things like numbers and arrow. The result is that keybindings that
should only have been added for keycodes from the first level, are
replaced by some unexpected keycode where the same keysym was found on
another level.
An example of this is the up-arrow key and l symbol. Normally you'd find
both the up-arrow symbol and the l symbol on the first level and be done
with it. However, on the German Neo-2 layout, layout level 4 maps the
KEY_E to the l symbol, while layout level 4 maps KEY_E to up-arrow.
Which ever gets to take priority is arbitrary, but for this particular
case KEY_E incorrectly mapped to up-arrow instead of the l symbol,
causing the keyboard shortcut Super+l, which would normally lock the
screen, to trigger the workspace-up (Super+up-arrow) key binding.
https://bugzilla.gnome.org/show_bug.cgi?id=789300
This protocol is limited to Xwayland only and is not visible/usable by
any other client.
Mutter uses the following mechanisms to determine if an X11 client
should be granted a grab:
- is "xwayland-allow-grabs" set?
- if set, is the client blacklisted?
- otherwise, has the client set the X11 window property
_XWAYLAND_MAY_GRAB_KEYBOARD on the window using a client message?
- if not, is it a client white-listed either via the default system
list or the settings "xwayland-grab-access-rules"?
https://bugzilla.gnome.org/show_bug.cgi?id=783342
Add a new client message "_XWAYLAND_MAY_GRAB_KEYBOARD" that X11 clients
can use to tell mutter this is a well behaving X11 client so it may
grant the keyboard grabs when requested.
An X11 client wishing to be granted Xwayland grabs by gnome-shell/mutter
must send a ClientMessage to the root window with:
- message_type set to "_XWAYLAND_MAY_GRAB_KEYBOARD"
- window set to the xid of the window on which the grab is to be issued
- data.l[0] to a non-zero value
Note: Sending this client message when running a plain native X11
environment would have no effect.
https://bugzilla.gnome.org/show_bug.cgi?id=783342
MetaWindowXwayland derives from MetaWindowX11 to allow for some Xwayland
specific vfunc that wouldn't apply to plain X11 windows, such as
shortcut inhibit routines.
https://bugzilla.gnome.org/show_bug.cgi?id=783342
The xdg-output protocol aims at describing outputs in way which is
more in line with the concept of an output on desktop oriented systems.
For now it just features the position and logical size which describe
the output position and size in the global compositor space.
This is however much useful for Xwayland to advertise the output size
and position to X11 clients which need this to configure their surfaces
in the global compositor space as the compositor may apply a different
scale from what is advertised by the output scaling property (to achieve
fractional scaling, for example).
This was added in wayland-protocols 1.10.
https://bugzilla.gnome.org/show_bug.cgi?id=787363
When the top window actor is destroyed, we need to make sure that
all its references are removed or it could be picked again in next
windows sync, causing crashes.
Since the window might or might not be destroyed when removed (depending
weather animations are in progress over it or not), it's just safer
to wait it to be destroyed before cleaning up any of its reference.
https://bugzilla.gnome.org/show_bug.cgi?id=791006
Changing the test monitor managers ability to rotate CRTCs in one test
affected the next test. Avoid leaking such state by resetting it before
each test. To continue passing, some tests needed to be updated
regarding to still pass.
https://bugzilla.gnome.org/show_bug.cgi?id=788915
Add a test case that checks that we don't try to revert to a
laptop-panel-only configuration after closing the lid after an external
monitor is connected.
https://bugzilla.gnome.org/show_bug.cgi?id=788915
We only counted configured monitors and whether the config was
applicable (could be assigned), howeverwe didn't include disabled
monitors when comparing. This could caused incorrect configurations to
be applied when trying to use the previous configuration.
One scenario where this happened was one a system with one laptop
screen and one external monitor that was hot plugged some point after
start up. When the laptop lid was closed, the 'previous configuration'
being the configuration where only the laptop panel was enabled, passed
'is-complete' check as the number of configured monitors were correct,
and the configuration was applicable.
Avoid this issue by simply comparing the configuration key of the
previous configuration and the configuration key of the current state.
This correctly identifies a laptop panel with the lid closed as
inaccessible, thus doesn't incorrectly revert to the previous
configuration.
https://bugzilla.gnome.org/show_bug.cgi?id=788915
When deriving the list of disabled monitors when creating new monitors
configs, don't include the laptop panel if the lid is currently closed,
as we consider the laptop panel nonexistent when the laptop lid is
closed when it comes to configuration.
The laptop panel connector(s) will either way be appropriately disabled
anyway, as the field listing disabled monitors in the configuration do
not affect actual CRTC/connector assignments.
https://bugzilla.gnome.org/show_bug.cgi?id=788915
If a parent doesn't have a window, it means it could have been
dismissed (for example due to a input serial race), but the more recent
popup might win the input serial race and try to map anyway. This would
result in a crash later on when trying to process the placement rule,
as the parent already has no window.
https://bugzilla.gnome.org/show_bug.cgi?id=790358
Move the top-most-popup correctness check to the finish_popup_setup()
function after checking the serial. If we pass the serial check, we
should have reached a state that if there are any popups they should be
the one from the same client.
Also avoid failing a client that correctly set the top-most popup at map
time, but where at the time of processing the top most popup have
already been dismissed by the compositor for some arbitrary reason.
https://bugzilla.gnome.org/show_bug.cgi?id=790358
Commit b1a0bf891 broke the previous logic that we would only fallback
to the root cursor if 1) windows are not interactable or 2) no window
cursor is currently set (i.e. not hovering over any window). Now it
will set up the root cursor if it's NULL, which breaks clients
explicitly setting an invisible cursor. This commit restaurates the
previous behavior.
https://bugzilla.gnome.org/show_bug.cgi?id=754806
This function is supposedly not failable, so just move the theme_dirty
flag clearing to the beginning of the function. Protects against cases
where requesting a cursor image may result in it being loaded and set
as a texture, which emits ::texture-changed, which may end up requesting
the cursor image again.
https://bugzilla.gnome.org/show_bug.cgi?id=754806
As wayland implements the cursor role, it consists of a persistent
MetaCursorSprite that gets the backing texture changed. This is
inconvenient for the places using MetaCursorTracker to track cursor
changes, as they actually track MetaCursorSprites.
This signal will be used to trigger emission of
MetaCursorTracker::cursor-changed, which will make users able to
update accordingly.
https://bugzilla.gnome.org/show_bug.cgi?id=754806
Just like X11/XFixes behaves, the current cursor is not affected
by its visibility, so it can be queried while invisible (possibly
to be replaced).
For this, keep an extra effective_cursor pointer that will be
either equal to displayed_cursor (maybe a bit of a misnomer now)
or NULL if the cursor is invisible. The MetaCursorRenderer
management is tied to the former, and the ::cursor-changed signal
emission to the latter.
https://bugzilla.gnome.org/show_bug.cgi?id=754806
We must emit ::dnd-leave to pair the ::dnd-enter that shall be
emitted whenever the plugin grab begins, otherwise we leave
listeners unable to clean up if the plugin begins and ends a
grab while there is an ongoing DnD operation.
https://bugzilla.gnome.org/show_bug.cgi?id=784545
MetaWaylandKeyboard maintains its own xkb_state used to update Wayland
clients.
Add the necessary hooks to make sure the sticky keys modifier masks set
in clutter-evdev are also applied in MetaWaylandKeyboard's xkb_state so
that Wayland clients also benefit from sticky keys.
https://bugzilla.gnome.org/show_bug.cgi?id=788564
We tried to get the geometry scale, which may depend on the main
logical monitor assigned to the window. To avoid dereferencing a NULL
logical monitor when headless, instead assume the geometry scale is 1.
https://bugzilla.gnome.org/show_bug.cgi?id=788764
Proprietary drivers such as ARM Mali export EGL_KHR_platform_gbm instead
of EGL_MESA_platform_gbm. As such, GBM platform check should be done for
both MESA and non-MESA drivers.
https://bugzilla.gnome.org/show_bug.cgi?id=780668
Bluetooth mouse usually goes in sleep state after a timeout, when that
happen the mouse is disconnected and on_device_removed function is
called. Before the patch if a touch device is available the
on_device_removed function hide the cursor. The issue is that the cursor
does not reappear once the bluetooth mouse is reconnected because
MetaBackend::current_device_id is not invalidated when on_device_removed
was called.
The patch set MetaBackend::current_device_id to 0 if the current device
is removed. This will make update_last_device to be triggered as soon as
another input device is used or the bluetooth mouse reconnect, as
consequence that the cursor reappear. The id 0 is never given to devices
and can safely used as undefine id.
https://bugzilla.gnome.org/show_bug.cgi?id=761067
The DRM properties container must be destroyed with
drmModeFreeObjectProperties, and the connectors must be freed on every
caller. Also make it sure that gbm_device structs are destroyed with the
MetaRendererNativeGpuData that owns them.
https://bugzilla.gnome.org/show_bug.cgi?id=789984
This is not a leak per se, but it seems too easy to make valgrind
SIGSEGV due to MetaBackground disconnecting signals from an already
destroyed MetaScreen when trying to SIGTERM gnome-shell. Keeping a
reference fixes this.
https://bugzilla.gnome.org/show_bug.cgi?id=789984
The layout group determines what actual keyboard layout in the keymap
to use when translating modifier state and key codes to key syms.
When changing a keymap to another, the layout groups has no relation to
the layout groups in the old keymap, thus there is no reason to
transfer it to the new state.
This fixes an issue where the xkb state in meta-wayland-keyboard.c got
desynchronized with the xkb state in clutter-device-manager-evdev.c.
https://bugzilla.gnome.org/show_bug.cgi?id=789300
On some architectures, including both GLES3/gl3.h GL/gl.h will cause
compilation issues due to incompatible type definitions. To avoid
running into that issue while building on other architectures, make
sure we haven't included GL/gl.h by accident.
https://bugzilla.gnome.org/show_bug.cgi?id=788695
The handlers depend on a role being assigned. Destroying the window
causes it to become unmapped, which would sometimes trigger one of the
handlers, resulting in an is-assigned assert hitting in one of the
handlers. Avoid this by disconnecting the handlers earlier, so that
there is no risk that any them being triggered before the role is
assigned.
https://bugzilla.gnome.org/show_bug.cgi?id=789552
In the unlikely case that a surface is moved by the compositor while
holding a pointer confinement, we also need to update the pointer
position when the surface actor gets moved.
https://bugzilla.gnome.org/show_bug.cgi?id=782344
Both notify::position on the surface actor and position-changed on
MetaWindow are listened to, in order to trigger wl_output updates for
wl_surfaces whenever the surfaces move across them.
Both signals are necessary in order to cater for toplevel and subsurface
relocations (Because it's the parent window actor what changes position
in this last case).
Also, shuffle signal disconnection, so each signal goes away with
the object reference held by MetaWaylandSurface.
https://bugzilla.gnome.org/show_bug.cgi?id=782344
The org.gnome.desktop.peripherals.trackball.scroll-wheel-emulation-button
setting contains buttons X11-style. Work out the BTN evcode that applies
to it when applying the setting on the libinput device.
https://bugzilla.gnome.org/show_bug.cgi?id=787804
It looks that there are some extensions that run a Mainloop on startup,
causing to dispatch a clutter paint before the compositor is even available.
In such scenario a MetaWindow could try to start a simple effect
using a compositor plugin which is not there yet.
Then in order to catch these bugs we can now assert that the expected
conditions are valid, so that gnome-shell will provide a dumpstack to
debug the real offending JS code.
https://bugzilla.gnome.org/show_bug.cgi?id=789223
There are cases when no compositor is available (yet) but a MetaWindow tries
to start a simple effect using a compositor plugin which is not available.
In that case we should just ignore any request and protect ourselves from
crashes.
https://bugzilla.gnome.org/show_bug.cgi?id=789223
This function might be called by components with invalid plugin manager
(as it might happen to MetaWindow when the compositor isn't initialized
properly), so we need to protect ourselves from crashes.
https://bugzilla.gnome.org/show_bug.cgi?id=789223
Check that if there are multiple modes with the same ID (resolution,
refresh rate and handled flags) we correctly add the preferred mode to
the list of monitor modes.
https://bugzilla.gnome.org/show_bug.cgi?id=789153
When generating MetaMonitorMode's, prefer CRTC modes that has the same
set of flags as the preferred mode. This not only is probably a better
set of configurable modes, but it'll guarantee that the preferred mode
is added.
This fixes a crash when the preferred mode was not the first mode with
the same resolution, refresh rate and set of handled modes.
https://bugzilla.gnome.org/show_bug.cgi?id=789153
Under X11 we can only ever have the same scale configured on all
monitors. In order to use e.g. scale 2 when there is a HiDPI monitor
connected, we must not disallow it because there is a monitor that does
not support scale 2. Thus we must show the same scale for every monitor
and monitor mode, even though it might result in a bad experience.
Do this by iterating through all the monitors adding all supported
scales by the preferred mode, combining all the supported scales. This
supported scales list is then used for all monitor and modes no matter
what.
https://bugzilla.gnome.org/show_bug.cgi?id=788901
When determining whether we should unredirect a window or not, ignore
offscreen windows, and just check the top most visible window.
Previously this was not an issue, but since 'stack-tracker: Keep
override redirect windows on top' we started sorting the UI frames
window, which is an offscreen override redirect window, on top, causing
the unredirect checking code to always check whether to unredirect the
UI frames window. This effectively disabled the compositor bypass
functionality.
https://bugzilla.gnome.org/show_bug.cgi?id=788493
Adding an internal signal and use it to update the internal state before
emitting "monitors-changed" which will be repeated by the screen to the world.
https://bugzilla.gnome.org/show_bug.cgi?id=788860
Don't use MAX(logical monitor scales) to determine the UI scaling
factor, just use the primary logical monitor. That's where the shell UI
will most likely be.
https://bugzilla.gnome.org/show_bug.cgi?id=788820
When we received two hot plug events that both resulted in headless
configuration, we tried to find a new window monitor given the old.
That resulted in a null pointer dereference; avoid that by only trying
to find the same monitor if there was an old one.
https://bugzilla.gnome.org/show_bug.cgi?id=788607
On hybrid GPU systems, hardware cursors needs to be realized on all the
GPUs, as scanout cursor planes cannot be shared. Do this by moving gbm
buffer and drm buffer ID management to a per GPU struct, realizing a
cursor on each GPU when previously only realized on the primary GPU.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
A hybrid GPU system is a system where more than one GPU is connected to
connectors. A common configuration is having a integrated GPU (iGPU)
connected to a laptop panel, and a dedicated GPU (dGPU) connected to
one or more external connector (such as HDMI).
This commit adds support for rendering the compositor stage using the
iGPU, then copying the framebuffer content onto a secondary framebuffer
that will be page flipped on the CRTC of the dGPU.
This can work in two different ways: GPU accelerated using Open GL ES
3, or CPU unaccelerated.
When supported, GPU accelerated copying works by exporting the iGPU
onscreen framebuffer as a DMA-BUF, importing it as a texture on a
separate dGPU EGL context, then using glBlitFramebuffer(), blitting it
onto a framebuffer on the dGPU that can then be page flipped on the dGPU
CRTC.
When GPU acceleration is not available, copying works by creating two
dumb buffers, and each frame glReadPixels() from the iGPU EGL render
context directly into the dumb buffer. The dumb buffer is then page
flipped on the dGPU CRTC.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Add helper functions and macros for managing and drawing OpenGL ES 3.
It will be used for blitting framebuffers between multiple GPUs in
hybrid GPU systems.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Eventually, we'll render buffers without using Cogl, and for this we
need to be able to do things like creating, destroying and changing the
context, as well as swapping buffers.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
First find the primary GPU and open it. Then go through all other
discovered GPUs with connectors and add those too. MetaRendererNative
still fails to initialize when multiple added GPUs and
MetaCursorRendererNative still always falls back on OpenGL based cursor
rendering when there are multiple GPUs.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
When creating a renderer with a custom winsys (which is always how
mutter uses cogl) make it possible to pass a user data with the winsys.
Still unused.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Make dumb buffer creation/destruction reusable by introducing a
MetaDumbBuffer type (private to meta-renderer-native.c). This will
later be used for software based fallback paths for copying render GPU
buffers onto secondary GPUs.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
On a CRTC that doesn't report any transforms at all, setting the normal
transform will fail. Avoid failing by checking if any transforms are
supported before applying it, and early out if no transforms are
supported.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Get rid of some technical dept by removing the support in the native
backend for drawing the the whole stage to one large framebuffer.
Previously the only way to disable stage views was to set the
MUTTER_STAGE_VIEWS environment variable to 0; doing that now will cause
the native backend to fail to initialize.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Don't permanently fall back to OpenGL based cursor rendering when
setting the HW cursor fails with EACCES as that may happen on VT
switching and other things temporarily revoking fd access.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
When drmHandleEvent() returns an error and errno is set to EAGAIN,
instead of ending up in a busy loop, poll() the fd until there is
anything to read.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
The prefix, if any, of a variable name often contains information about
the namespace (such as clutter_backend is the ClutterBackend, while
backend is a MetaBackend). Clean up some more inconsistencies in
meta-renderer-native.c where various variable names were egl_ prefixed
but in fact was Cogl types.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
In order to eventually support multilpe GPUs with their own connectors,
split out related meta data management (i.e. outputs, CRTCs and CRTC
modes) into a new MetaGpu GObject.
The Xrandr backend always assumes there is always only a single "GPU" as
the GPU is abstracted by the X server; only the native backend (aside
from the test backend) will eventually see more than one GPU.
The Xrandr backend still moves some management to MetaGpuXrandr, in
order to behave more similarly to the KMS counterparts.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Pass the backend to a new factory function, and keep a pointer to the
monitor manager, which is accessed elsewhere in the same file instead of
fetching the singleton. The HW cursor initialization part is also made
more obvious, without depending on seemingly irrelevant clutter
features.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Move finding, opening and managment of the KMS file descriptor to
MetaMonitorManagerKms. This means that the monitor manager creation can
now fail, both if more than one GPU with connectors is discovered, or
if finding or opening the primary GPU fails.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
The monitor manager instance was created and setup in one step; at
construction. This is problematic if, in the future, the monitor manager
creation can fail, as the monitor manager is created quite late.
To make it possible to in the future fail creating a monitor manager,
create the instance very early when initiating the backend, then on
post init backend setup, "setup" the monitor manager state, i.e. read
the current state and setup the stage.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
The error was printed, then dropped, eventually resulting in another
generic error being printed. Lets just propogate the error all the way
up instead.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Move code dealing with Xrandr MetaCrtcs and related functionality to its
own file. Eventually, MetaCrtcCrtc should be introduced, based on
MetaCrtc, and this commit is in preparation for that.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Move code dealing with X11 MetaOutputs and related functionality to its
own file. Eventually, a MetaOutputXrandr should be introduced, based on
MetaOutput, and this commit is in preparation for that.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Move code dealing with MetaCrtcKms and related functionality to its
own file. Eventually, MetaCrtcKms should become a GObject based on
MetaCrtc, and this commit is in preparation for that.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Instead of passing it around or fetching the singleton, keep a pointer
to the monitor manager that owns the CRTC. This will eventually be
replaced with a per GPU/graphics card object.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Move code dealing with MetaOutputKms and related functionality to its
own file. Eventually, MetaOutputKms should become a GObject based on
MetaOutput, and this commit is in preparation for that.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Instead of passing it around or fetching the singleton, keep a pointer
to the monitor manager that owns the output. This will eventually be
replaced with a per GPU/graphics card object.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Convert MetaCrtcMode from a plain struct to a GObject. This changes the
storage format, and also the API, as the API was dependent on the
storage format.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Turn MetaCrtc into a GObject and move it to a separate file. This
changes the storage format, resulting in changing the API for accessing
MetaCrtcs from using an array, to using a GList.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Add meta_logical_monitor_foreach_crtc() helper to iterate over all the
active CRTCs driving the monitors associated with the specified logical
monitor.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
Turn MetaOutput into a GObject and move it to a separate file. This
changes the storage format, resulting in changing the API for accessing
MetaOutputs from using an array, to using a GList.
https://bugzilla.gnome.org/show_bug.cgi?id=785381
The modifier event was only added in v3 of the client; sending it to
older clients (e.g. GStreamer waylandsink) causes them to disconnect
immediately.
Send the older 'format' event to all clients, and only send the newer
'modifier' event to resource versions 3 or above.
https://bugzilla.gnome.org/show_bug.cgi?id=788558
If a configuration key matched a current system state, but no monitor
mode was found (for example because of an incorrect refresh rate),
discard it while logging a warning.
https://bugzilla.gnome.org/show_bug.cgi?id=787668
People that relied on xsetwacom to configure their tablets used to get
away with this by disabling the wacom g-s-d plugin prior to running
their scripts. This is not possible anymore with mutter managing device
configuration.
Given that X11 shall not go away soon and there's a core of stubbornly
accustomed users, provide a MUTTER_DISABLE_WACOM_CONFIGURATION envvar
to provide *some* way to do this.
To keep feature parity with the Wayland backend, and
to improve the overall tiling experience with GTK apps,
add the _GTK_EDGE_CONSTRAINTS X11 atom and update it
when necessary.
https://bugzilla.gnome.org/show_bug.cgi?id=751857
Following up the previous patch, this patch makes the
Wayland backend send the edge constraints through a
custom protocol extension internal to GTK.
As it mature, we can think of upstreaming the protocol
to Wayland itself.
https://bugzilla.gnome.org/show_bug.cgi?id=751857
GTK has the ability to handle client-decorated windows
in such a way that the behavior of these windows must
match the behavior of the current window manager.
In Mutter, windows can be tiled horizontally (and, in
the future, vertically as well), which comes with a few
requirements that the toolkit must supply. Tiled windows
have their borders' behavior changed depending on the
tiled position, and the toolkit must be aware of this
information in order to properly match the window manager
behavior.
In order to provide toolkits with more precise and general
data regarding resizable and constrained edges, this patch
makes MetaWindow track its own edge constraints.
This will later be used by the backends to send information
to the toolkit.
https://bugzilla.gnome.org/show_bug.cgi?id=751857
When computing a potential match for a tiled window, there is a
chance we face the case where 2 windows really complement each
other's tile mode (i.e. left and right) but they have different
sizes, and their borders don't really touch each other.
In that case, the current code would mistakenly assume they're
tile matches, and would resize them with either a hole or an
overlapping area between windows. This is clearly a misbehavior
that is a consequence of the previous assumptions pre-resizable
tiles.
This patch adapts the tile match algorithm to also consider the
touching edges when computing the matching tile, unless:
* the window is not currently tiled (for example when computing
the tile preview)
* the window is currently resized in tandem with an existing
tile match
https://bugzilla.gnome.org/show_bug.cgi?id=645153
bar
When a pair of tiled windows are grouped together, they
are treated as parts of a whole and interacting with one
affects the other.
Following the idea that sibling tiled windows are treated
as part of the same group, they should also be raised and
lowered together.
It is still possible to break tiled windows grouping by
simply untiling the window with the keyboard or by grabbing
and resizing or moving the window with the cursor.
This patch makes sibling tiled windows be lowered and raised
in tandem. For future reference, this behavior is documented
in [1].
[1] https://wiki.gnome.org/GeorgesNeto/MinutesOfFeaneron/Tilinghttps://bugzilla.gnome.org/show_bug.cgi?id=645153
There is a variable in meta_window_edge_resistance_for_resize
that isn't really helpful: it just assumes TRUE, and is passed
to apply_edge_resistance_to_each_side.
This patch removes that useless variable and simply pass TRUE
instead.
https://bugzilla.gnome.org/show_bug.cgi?id=645153
When windows are tiled, it improves the interaction with
them when they have a set of snapping edges relative to
the monitor. For example, when there's a document editor
and a PDF file opened, I might want to rescale the former
to 2/3 of the screen and the latter to 1/3.
These snapping sections are not really tied to any other
window, and only depend on the current work area of the
window. Thus, it is not necessary to adapt the current
snapping edge detection algorithm.
This patch adds the necessary code in edge-resistance.c
to special-case tiled windows and allow them to cover
1/4, 1/3 and 1/2 (horizontally) of the screen. These
values are hardcoded.
https://bugzilla.gnome.org/show_bug.cgi?id=645153
After the introduction of the possibility to resize tiled windows,
it is a sensible decision to make windows aware of their tiling
match. A tiling match is another window that is tiled in such a
way that is the complement of the current window.
The newly introduced behavior attepts to make tiling as smooth as
possible, with the following rules:
* Windows now compute their tile match when tiling and, if there's
a match, they automatically complement the sibling's width.
* Resizing a window with a sibling automatically resizes the sibling
too to be the complement of the window's width.
* It is not possible to resize below windows' minimum widths.
https://bugzilla.gnome.org/show_bug.cgi?id=645153
Now that tiled windows are resizable, the user may grow a tiled
windows until it covers the entire work area. As this makes the
window state mostly indistinguishable from maximization, avoid
subtle differences by properly maximizing the window in that case.
https://bugzilla.gnome.org/show_bug.cgi?id=645153
Currently tiled windows are not resizable and their size is fixed
to half the screen width. Adjust the code to work with fractions
other than half, and allow users to adjust the split by dragging
the window edge that is not constrained by a monitor edge.
Follow-up patches will improve on that by resizing neighboring
tiled windows by a shared edge, and making the functionality
available to client-side decorated windows implementing the
new edge constraints protocol.
Now that the preview tile mode has been split from the window's
tile_mode property, it is much more natural to pass the requested
tile_mode to the tile() function instead of setting it externally
and calling the function to apply the state.
The existing semantics of the tile_mode property are terribly confusing,
as it depends on some other property whether it represents the requested
or current mode. Clear this up by just using separate variables for the
two. As it is unlikely that we will ever support more than one tile
preview, we can track the requested mode globally instead of adding
another per-window variable.
https://bugzilla.gnome.org/show_bug.cgi?id=645153
update_num_workspaces() is a no-op when the number of workspaces
did not actually change. That is fine, except that we still want
to initialize the _NET_NUMBER_OF_DESKTOPS hint on startup to not
break components like nautilus-desktop that rely on it.
https://bugzilla.gnome.org/show_bug.cgi?id=760651
Automatic maximization is done when a window is almost the size of the
work area of a monitor. This makes no sense to try when there is no
monitor available, so skip trying to do this when headless.
https://bugzilla.gnome.org/show_bug.cgi?id=787637
When we are headless, treat this as if the window is always not monitor
sized. This might cause windows to temporarly become redirected while
being headless, but this is harmless as when a new monitor is
connected, we'll recalculate weather it should be redirected or not.
https://bugzilla.gnome.org/show_bug.cgi?id=787637
Check that configurations where monitors are disabled are properly
used. Also test that old configurations with explicitly disabled
outputs are migrated properly.
https://bugzilla.gnome.org/show_bug.cgi?id=787629
When saving and restoring monitor configurations, we must take disabled
monitors into account, as otherwise one cannot store/restore a
configuration where one or more monitors are explicitly disabled. Make
this possible by adding a <disabled> element to the <configure> element
which lists the monitors that are explicitly disabled. These ones are
included when generating the configuration key, meaning they'll be
picked up correctly.
https://bugzilla.gnome.org/show_bug.cgi?id=787629
When dynamic workspaces are used, we pick up an existing NUMBER_OF_DESKTOP
hint in meta_screen_init_workspaces() to properly restore workspaces
on restart. Unconditionally setting that hint to 1 *before* reading it
breaks that, and we end up shifting all windows to the first workspace.
This reverts commit 8532b10290.
https://bugzilla.gnome.org/show_bug.cgi?id=760651
Also adds a soft assert to meta_window_is_on_primary_monitor() to make
it easier to spot when callers might want to handle headless
in a certain way.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
The XIQueryDevice function used by device_query_area can return a NULL
pointer and set n_devices to a negative number in some cases. We add
additional checks to prevent a segfault.
https://bugzilla.gnome.org/show_bug.cgi?id=787649
For now we abuse of meta_window_get_flatpak_id not to break the APIs,
so that it's working seamlessly in gnone shell too.
Rename flatpak_id to sandboxed_app_id internally to get prepared to the future
API.
https://bugzilla.gnome.org/show_bug.cgi?id=788217
If a non-latin based keyboard layout is active, for example Cyrillic,
keybindings won't work unless we resolve the bound keysyms on a
secondary latin based layout. So, to make keybindings work on non-latin
based layouts, detect if a keymap doesn't have all of the basic latin
letters (a-z) and resolve from an additional US layout as well.
https://bugzilla.gnome.org/show_bug.cgi?id=787016
Add the infrastructure to resolve keybinding symbols from multiple
layouts. It is still unused, but will be, when the primary layout does
not have the required latin keysyms in it.
https://bugzilla.gnome.org/show_bug.cgi?id=787016
When rotating 90/270 degrees we need to swap width and height. This fixes
the screen going black and the following errors showing in the journal:
gnome-shell[1097]: Failed to set CRTC mode 800x1280: No space left on device
gnome-shell[1097]: Failed to flip: Device or resource busy
gnome-shell[1097]: Failed to set CRTC mode 800x1280: No space left on device
gnome-shell[1097]: Failed to set CRTC mode 800x1280: No space left on device
When rotating a tablet with accelerometer 90/270 degrees.
https://bugzilla.gnome.org/show_bug.cgi?id=787836
Make sure to call set_number_of_spaces_hint in meta_screen_new.
_NET_NUMBER_OF_DESKTOPS is required by nautilus-desktop to correctly
get the desktop workarea.
https://bugzilla.gnome.org/show_bug.cgi?id=760651
We currently have a hard coded limit on logical monitor sizes, meant
for filtering out monitor scales that would result in awkward desktop
sizes. This has the side effect of also disqualifying scale 1 for
resolutions that themself are lower than the mentioned limit. To avoid
listing no supported scales, always add the fallback scale 1 if no
other was added.
https://bugzilla.gnome.org/show_bug.cgi?id=787477
When we update state, we might not have set the current config yet (for
example if the Xrandr assignment didn't change), so pass the monitors
config we should derive from instead of fetching it from the monitor
config manager.
https://bugzilla.gnome.org/show_bug.cgi?id=787477
We currently only save synchronously when running the test suite, but
should still not leak the generated config buffer. We also created the
cancellable but never used it if we saved synchronously, so lets stop
doing that too.
https://bugzilla.gnome.org/show_bug.cgi?id=787477
The cancellable should only be cleared if we weren't cancelled, as if
we were cancelled, the path cancelling have already cleared the
cancellable.
https://bugzilla.gnome.org/show_bug.cgi?id=787477
We cleaned up an unused monitor config list, but what we should do is
clear up the logical monitor config list. This commit does that, as
well as removes the unused monitor config list.
We have not enough control over the sources of the refresh rate
float variable to make == comparisons reliable, add some room
when comparing these.
https://bugzilla.gnome.org/show_bug.cgi?id=787668
We would free the shortcut inhibit data only when the client destroys
its request, which is not the case when the clients itself is
destroyed, leading to a leak of the shortcut inhibit data.
Free the data on resource destruction instead, and simply destroy the
resource on destroy request.
https://bugzilla.gnome.org/show_bug.cgi?id=787568
On Wayland, the grab()/ungrab() in gtk+/gdk are wired to the shortcut
inhibitor mechanism, which in turn shows the dialog, which can take
focus away from the client window when the dialog is shown.
If the client issues an ungrab() when the keyboard focus is lost, we
would hide the dialog, causing the keyboard focus to be returned to the
client surface, which in turn would issue a new grab(), so forth and so
on, causing a continuous show/hide of the shortcut inhibitor dialog.
To avoid this issue, keep the dialog around even if the shortcut inhibit
is canceled by the client, so that the user is forced to make a choice
that we can reuse on the next request without showing the dialog again.
Instead of hiding the dialog when the shortcut inhibitor is destroyed by
the client, we simply mark the request as canceled and do not apply the
user's choice.
https://bugzilla.gnome.org/show_bug.cgi?id=787568
The brightness is about the vignette. Add a 'vignette_' prefix in order
to give more context. Keep the property name as it is, doing so we don't
break any plugin (e.g., gnome-shell).
https://bugzilla.gnome.org/show_bug.cgi?id=786618
Add the word 'vignette' in order to give some context on what the
snippets do. This will be useful later when we land other snippets
for the gradient effect.
https://bugzilla.gnome.org/show_bug.cgi?id=786618
The reverted commit seems to cause
https://bugzilla.gnome.org/show_bug.cgi?id=787240 for some reason. Lets
be safe and revert it for now, as the code freeze is just around the
corner.
This partly (it doesn't reintroduce a whitespace issue) reverts commit
dbc63430d8.
When cleaning up the display name string management, the display name
string retrieved from libwayland-server was also passed to free() on
clean up. This is invalid as the display name string ownership is not
transferred to us. Fix this by strdup:ing the string before saving it.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
Give clients (such as Xwayland) a chance to bind the wl_output global
before we continue, otherwise there is an significant risk that mutter
won't see the bind request until after the next hot plug which might
have destroyed the global object.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
This avoids updating state (such as position, size etc) when going
headless. Eventually, when non-headless, things will be updated again,
and not until then will we be able to update to a valid state.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
In order to give the clients the best chance to bind the wl_output
before we later remove it (for example on fast hot plugs or in the test
suite), flush the client sockets after creating the global.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
The foreach CRTC monitor mode helper incorrectly iterated over outputs
without CRTC when non-tiled modes were set on tiled monitors. This was
not expected by callers, so fix the helper to only iterate over active
outputs (that has or should have a CRTC).
The test cases uses the incorrect behaviour of the foreach CRTC helper
to check that the disabled outputs mode are set to NULL, so add a
foreach output helper and change the tests to use that instead.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
If there are no active logical monitors, don't try to dereference a
NULL one to get a preferred output winsys id. Instead just set an
invalid one.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
When headless, we don't have any logical monitors to derive a screen
size from, but we can't set it to empty as that will cause issues with
the clutter stage, UI widget layout and other things. To avoid such
issues, just fall back to a 640 x 480 screen size when headless.
https://bugzilla.gnome.org/show_bug.cgi?id=730551
Due to rounding issues, we can't assume a floating point calculation
will end up on an integer, even if we got the factor from the reverse
calculation. Thus, to avoid casting away values like N.999... to N,
when they should really be N+1, round the resulting floating point
calculation before casting it to int.
This fixes an issue where using the scale ~1.739 on a 1920x1080 mode
resulted in error when setting the mode, as the calculated size of the
framebuffer was only 1919x1080.
https://bugzilla.gnome.org/show_bug.cgi?id=786918
When a screen cast session is stand-alone, i.e. not created given a
remote desktop session managing it, allow calling the Start/Stop
methods to start and stop it.
https://bugzilla.gnome.org/show_bug.cgi?id=784199
