We currently calling gdk_display_map_keyval up to
once per key event per shortcut trigger, and that function
does an expensive loop over the entire keymap and
allocates an array. Avoid this by caching the entries
in a single array, and have a lookup table for finding
the entries for a keyval.
To do this, change the GdkKeymap.get_entries_for_keyval
signature, and change the ::keys-changed signal to be
RUN_FIRST, since we want to clear the cache in the class
handler before running signal handlers. These changes are
possible now, since keymaps are no longer public API.
The api to configure surfaces is now GdkToplevelLayout
and GdkPopupLayout. Unfortunately, there's still quite
a bit of internal use of GdkGeometry that will take some
time to clean up, so move it go gdkinternals.h for now.
GdkEvent has been a "I-can't-believe-this-is-not-OOP" type for ages,
using a union of sub-types. This has always been problematic when it
comes to implementing accessor functions: either you get generic API
that takes a GdkEvent and uses a massive switch() to determine which
event types have the data you're looking for; or you create namespaced
accessors, but break language bindings horribly, as boxed types cannot
have derived types.
The recent conversion of GskRenderNode (which had similar issues) to
GTypeInstance, and the fact that GdkEvent is now a completely opaque
type, provide us with the chance of moving GdkEvent to GTypeInstance,
and have sub-types for GdkEvent.
The change from boxed type to GTypeInstance is pretty small, all things
considered, but ends up cascading to a larger commit, as we still have
backends and code in GTK trying to access GdkEvent structures directly.
Additionally, the naming of the public getter functions requires
renaming all the data structures to conform to the namespace/type-name
pattern.
For the X11 backend, keep a list of monitors for which the surface
intersects the monitor area.
Whenever the X11 surface is configured, check against the list of
monitors to determine whether it enters a new monitor or if it left a
monitor, to emit the corresponding ::enter/leave-monitor signals just
like a Wayland compositor would.
As monitors can be added, removed or reconfigured at any time, redo
those checks whenever any of these events occur.
We don't need all of them, only the ones that contain public API. This
allows us to reduce the chance of a stray symbol getting incorrectly
added to the introspection data.
Add all of the keyboard translation results in the key event,
so we can translate the keyboard state at the time the event
is created, and avoid doing state translation at match time.
We actually need to carry two sets of translation results,
since we ignore CapsLock when matching accelerators, in
gdk_event_matches().
At the same time, drop the scancode field - it is only ever
set on win32, and is basically unused in GTK.
Update all callers.
On X11, there is no such equivalent to the inhibit shortcut protocol
found on Wayland.
To implement the inhibit_system_shortcuts API on X11, we emulate the
same behavior using grabs on the keyboard.
To avoid keeping active grabs on the keyboard that would affect other
X11 applications even when the surface isn't focused, the X11
implementation takes care of releasing the grabs as soon as the toplevel
loses focus.
Without this, the back buffers of the wrong size
keep being used, causing flickery misdraws, as
seen when expanding the expander in the popover
in widget-factory.
There is no shape combining going on anymore, so
call this just gdk_surface_set_input_region, and
remove the offset arguments too. All callers pass
0 anyway.
Update all callers and implementations.
Sprinkle various g_assert() around the code where gcc cannot figure out
on its own that a variable is not NULL and too much refactoring would be
needed to make it do that.
Also fix usage of g_assert_nonnull(x) to use g_assert(x) because the
first is not marked as G_GNUC_NORETURN because of course GTester
supports not aborting on aborts.
Drop the input-mode, since it only makes sense for
floating devices, which we don't have anymore. And renamt
::input-source to ::source, to match the getter.
Update all users.
replace all uses with const char * (non-interned).
Also remove a lot fo juggling from atom to GdkAtom to string and back.
The X Atom hash table is now mapping to (again, non-interned) strings.
We were not properly converting the coordinates we
got to root coordinates. This was showing up as offsets
between the actual drop target and the area where drops
can happen, e.g. when dragging over a stack switcher
to switch pages.
We were forgetting to clean up the ::xevent signal
handler in some error cases. Move the signal connection
later, when we know the drag is going forward, and
use g_signal_connect_object to make sure the signal
handler is not forgotten.
Restructure the getters for event fields to
be more targeted at particular event types.
Update all callers, and replace all direct
event struct access with getters.
As a side-effect, this drops some unused getters.
Make the event translator return a new event, instead of
filling in a half-constructed one.
Update the two implementation in GdkX11Display and
GdkDeviceManagerXI2.
Instead of passing a half-constructed event and expect
it to be filled in, pass the surface as in argument, and
add an out argument for a newly constructed GdkEvent.
Replace the gdk_surface_move_to_rect() API with a new GdkSurface
method called gdk_surface_present_popup() taking a new GdkPopupLayout
object describing how they should be laid out on screen.
The layout properties provided are the same as the ones used with
gdk_surface_move_to_rect(), except they are now set up using
GdkPopupLayout.
Calling gdk_surface_present_popup() will either show the popup at the
position described using the popup layout object and a new unconstrained
size, or reposition it accordingly.
In some situations, such as when a popup is set to autohide, presenting
may immediately fail, in case the grab was not granted by the display
server.
After a successful present, the result of the layout can be queried
using the following methods:
* gdk_surface_get_position() - to get the position relative to its
parent
* gdk_surface_get_width() - to get the current width
* gdk_surface_get_height() - to get the current height
* gdk_surface_get_rect_anchor() - to get the anchor point on the anchor
rectangle the popup was effectively positioned against given
constraints defined by the environment and the layout rules provided
via GdkPopupLayout.
* gdk_surface_get_surface_anchor() - the same as the one above but for
the surface anchor.
A new signal replaces the old "moved-to-rect" one -
"popup-layout-changed". However, it is only intended to be emitted when
the layout changes implicitly by the windowing system, for example if
the monitor resolution changed, or the parent window moved.
The returned position should be relative to the parent surface, but
GdkSurface::x,y were only managed properly for O-R windows. This makes
it correct for regular windows too.
The marks are averaged based on the name, so this makes more sense.
Also rename the map/unmap marks to have the same capitalization as
everything else.
When we use if (GDK_PROFILER_IS_RUNNING) this means we get an
inlined if (FALSE) when the compiler support is not compiled in, which
gets rid of all the related code completely.
We also expand to G_UNLIKELY(gdk_profiler_is_running ()) in the supported
case which might cause somewhat better code generation.
All the code in e.g. init_randr15() divides the physical resolutions with
the screen scale, however if we get the screen scale from xsettings
rather than e.g. GDK_SCALE the initial setup is using the wrong value.
So, whenever the screen scale size is changed we need to trigger
a re-read of the randr data
This adds a GDK_DEBUG=default-settings flag which disables reads
from xsettings and Xft resources, and enables this for the testsuite.
This is one less way to get different testresults depending on the
environment. In particular, it was failing the css tests for me
due to getting the wrong font size because i have a different dpi.
We only have implementations of this on X11 and Win32,
so make it available as backend api there.
Update all callers to use either the backend api, or
just monitor 0.
When a device is added, there are two references to it by the device
manager, the initial one and the one used for the id_table. Removing a
device only removed the reference added by the id_table resulting in the
GdkDevice being leaked.
https://gitlab.gnome.org/GNOME/gtk/merge_requests/1358
According to XDND "The XdndLeave message cancels the session.",
issue one when cancelling a drag, so the dest side has an opportunity
to forget about the GdkDrop.
The drag source might be cached and held alive, only disposed after
future drag begin operations. Ensure the drag surface gets hidden
properly or might might stay transparent but mapped till then.
Otherwise the icon "jumps" to the cursor position with its top left when
the animation starts.
This is especially visible if the dragged item is big, like when dragging
mails in Thunderbird.
We use a compilation symbol in our build to allow the inclusion of
specific headers while building GTK, to avoid the need to include only
the global header.
Each namespace has its own compilation symbol because we used to have
different libraries, and strict symbol visibility between libraries;
now that we have a single library, and we can use private symbols across
namespaces while building GTK, we should have a single compilation
symbol, and simplify the build rules.
The "iconified" state is mostly an X11-ism; every other platform calls
this state "minimized" because it may not involve turning a window into
an icon at all.
Previously, the manufacturer property of the GdkMonitor was NULL,
and having at least PNP id at GdkMonitor.manufacturer makes it
possible to distinguish between different monitors programmatically.
Windows/surface's aren't supposed to be explicitly moved by any external
part, so don't provide API for doing so. Usage throughout Gdk is
replaced by the corresponding backend variants.
The generic layer still does the heavy lifting, leaving the backends
more or less just act as thin wrappers, dealing a bit with global
coordinate transformations. The end goal is to remove explicit surface
moving from the generic gdk layer.
To separate how toplevels and popups are configured, a first step is to
introduce a resize-only vfunc for backends to implement. It's meant to
only configure toplevel windows, i.e. popups. Currently it's used for
both types, but introducing the resize-only API is a first step.
When unreffing the stream from a different thread, the close function
will schedule its cleanup asynchornously in the main thread.
We need to make sure the stream object stays alive for as long as
that hasn't happened, so ref() it.
Fixes#2003
To make a frame clock tick as long as any of the associated surfaces
expect to receive ticks, make the surfaces inhibit freezing the clock,
instead of directly tell the frame clock to freeze itself.
This makes it so that as long as any surface using a certain frame clock
is not frozen (e.g. just received a frame event from the display
server), the frame clock will not be frozen.
With this, the frame clock is initiated as frozen, and won't be thawed
until any surface inhibits freeze. It will be frozen again, when every
surface has that previously inhibited freeze uninhibited freeze.
If we set c_marshaller manually, then g_signal_newv() will not setup a
va_marshaller for us. However, if we provide c_marshaller as NULL, it will
setup both the c_marshaller (to g_cclosure_marshal_VOID__VOID) and
va_marshaller (to g_cclosure_marshal_VOID__VOIDv) for us.
The X backend was storing global coordinates
in surface->x/y, and keeping the parent-relative
positions in its own fields. Switch this around
to store the relative position in x/y, as is
expected by the frontend.
Now that popups share the frame clock of their
parent, we have to be much more careful about
freezing the clock, since that may stop updates
for another surface.
This commit makes two changes that make the
X11 handling of the frame clock more similar
to the Wayland backend:
- Use gdk_surface_freeze_updates instead of
gdk_surface_freeze_toplevel_updates to avoid
affecting the frame clock
- Bail out early in before_paint/after_paint
if the surface is frozen, to avoid affecting
the frame clock
Together, these two make the X11 popup surface
type work without freezing updates for the toplevel.
With separate clocks, the phases are not coordinated,
which messes with GTKs size allocation machinery treating
the entire widget tree as a whole, and causes us to
run into assertion where popups get drawn before they
are allocated.
Make them use o-r windows, and move
with their parent.
We do a sort-of ok job on stacking order
here - whenever the parent window gets a
ConfigureNotify, we just restack all popups
directly on top of their parent. This is good
enough to keep popups on top of their parent
while we drag it around, and it gets the popup
to disappear when raising another window on
top of the parent.
Store popup parents separately from transient-for
parents, since these are separate concepts with
different behaviors. And we need the parent in
the frontend, so we can use it in the fallback
move-to-rect implementation.
We don't need the complicated wrapper system anymore,
since client-side windows are gone. This commit moves
all the vfuncs to GtkSurfaceClass, and changes the
backends to just derive their surface implementation
from GdkSurface.
We want to use a gdk_surface_new_popup for popups,
and align the constructor names with the surface
types, so rename
gdk_surface_new_popup -> gdk_surface_new_temp
gdk_surface_new_popup_full -> gdk_surface_new_popup
The temp surface type will disappear eventually.
All the information in it is already contained
in the surface object we pass along, and none
of the backend implementations were using the
attributes at all.
We are not creating such surfaces anymore, and
they were only ever meaningfully implemented
on X11. Drop the concept, and the api for determining
if a surface is input-only.
We were adding incomplete frame timings to the
profile, which lead to occasional nonsense
numbers. Instead, only add timings to the profile
once we marked them as complete. This also
gives us an opportunity to add the presentation
time as a marker.
Besides requiring it at build time, require that the server the client
is running against exposes the XInput2 protocol. We no longer fallback
on a device manager for core events.
XInput2 is more than a decade old already, and the input improvements
there (and in every other backend really) make it untenable to have
support for X11 core input events dragging things behind.
The skip-taskbar, skip-pager and urgency hints were
only ever implemented for X11, and are not very useful
with modern desktops. Relegate the functionality to
x11 backend api, and drop the GtkWindow api.
Change the all the begin_drag and begin_move apis in
GdkSurface and GtkWindow to expect surface coordinates.
Update the x11 implementation to translate to root
coordinates where it matters. Wayland is ignoring the
coordinates anyway.
Some of the flags got lost in the meson transition or were demoted from
error flags to warning flags.
This commit reintroduces them.
It also includes fixes for the code that had warnings with those flags.
The big one being -Wshadow.
Tools on the same physical item have the same serial number, so the eraser
and the pen part of a single pen share that serial number. With the current
lookup code, we'll always return whichever tool comes first into proximity.
Change the code to use the hw id in addition to the serial number, this way we
can differ between two tools.
Generic tools (Bamboo, built-in tablets) always have the same serial number
assigned by the wacom driver. This includes the touch tool when the wacom
driver handles the touch evdev node (common where users require the wacom
gestures to work).
When the first device is the touch device, a tool is created with that serial.
All future tools now return the touch tool on lookup since they all share the
same serial number. Worse, this happens *across* devices, so the pen
event node gets assigned the touch tool because they all have the same serial.
Since we don't actually care about the touch as a tool, let's skip any unknown
tool. This captures pads as well.
Any wacom device currently sets the tool type to UNKNOWN. The wacom driver has
a property that exports the tool type as one of stylus, eraser, cursor, pad or
touch. Only three of those are useful here but that's better than having all
of them as unknown.
As per the spec:
> The back buffer can
> either be reported as invalid (has an age of 0) or it may be
> reported to contain the contents from n frames prior to the
> current frame.
So a buffer age of 1 means that the buffer was used in the last frame.
We were handling buffer_age==1 the same as buffer_age==0, i.e. we
returned the full damage for the surface.
[1] https://www.khronos.org/registry/EGL/extensions/EXT/EGL_EXT_buffer_age.txt
GtkEntryCompletion can rapidly release and claim ownership of the
primary selection. This generates multiple XFixesSelectionNotify events,
first stating that no one owns the selection, then another stating that
we own the selection. The notification that no one owns the selection
causes GtkEntryCompletion to deselect the text, breaking inline
autocompletion.
This fixes it by ignoring any XFixesSelectionNotify with a timestamp
earlier than our clipboard timestamp.
Fixes#14
Change GdkDrag::action to GdkDrag::selected-action, which is
more clearly different from actions, and follows the existing
name of the struct field and getter.
This lets us drop the ::action-changed signal for the
property change notification. But, can just as well move
the signal class handers which just update the cursor
to the ::action setter. No need to do this in the backends.
Rename gdkdnd.h to gdkdrag.h, to go along with gdkdrop.h
This commit includes the necessary updates to the X11, Wayland
and Broadway backends. Other backends have to be updated separately.
This is to go along with the newly introduced GdkDrop.
This commit includes the necessary updates to the X11, Wayland
and Broadway backends. Other backends have to be updated separately.
This might be foreign Windows and we don't want to create surfaces for
those.
Also, stop using GdkDragContext.dest_surface, that variable is meant to
go away.
In particular, this patch removes:
gdk_surface_get_events()
gdk_surface_set_events()
gdk_surface_get_device_events()
gdk_surface_set_device_events()
Event masks so far still exist for grabs.
GdkDragContext => GdkDrop
This is all in preparation of separation of the drag and drop.
Also, don't check for GDK_DRAG_PROTO_XDND anymore - it's the only
possible value for the protocol on the target side.
Use the new method of connecting to the xevent signal instead.
Also, don't consume the xevent, there might be other code listening for
it. And we don't use PropertyNotify in the generic code path anymore, so
it'll just be ignored there.
Instead of looking at the list of contexts, just look at the current
drop context. There is only one, after all.
Then remove the is_source argument from gdk_drag_context_find().
The filters now return TRUE/FALSE and no longer a GdkFilterReturn. They
also don't conform to the GdkFilterFunc typedef anymore but instead take
the arguments that they need.
In 01455399e8 ("gdk: do not deactivate surface on keyboard grabs"), we
made gdk avoid deactivating surfaces when another application takes a
keyboard grab, by using has_focus_window instead of has_focus. That however
broke activating surfaces when the gdk application acquired a grab itself,
in which case has_focus_window is false but has_focus is true.
We thus actually need to use both: surfaces should be activated either
because we have normal keyboard focus, or because we grabbed the keyboard.
This also renames HAS_FOCUS to APPEARS_FOCUSED to better reflect its
role.
Fixes#85
Includes implementation for Wayland and X11, which are the only backends
implementing the Startup Notification Protocol, returns NULL otherwise.
https://gitlab.gnome.org/GNOME/gtk/issues/1084
Similar to what has been done recently for DESKTOP_AUTOSTART_ID [1],
we need to get rid of this call to g_unsetenv() in the displays'
backends for X11 and Wayland, so that it's guarantee to happen any
thread is created, while still being accessible when needed.
Let's stash the value of this environment variable when loading the
GDK library, and provide a private method so that it can be retrieved
from the displays' backend when implementing gdk_display_make_default().
[1] https://gitlab.gnome.org/GNOME/gtk/commit/22269902
Closes: https://gitlab.gnome.org/GNOME/gtk/issues/979
The check survived from GTK2 when that function could still return
GdkPixmap and GdkFont objects and was accompanied by this comment:
/* We may receive events such as NoExpose/GraphicsExpose
* and ShmCompletion for pixmaps
*/
When pressing e.g. a window manager shortcut, which acquires keyboard grab,
Xorg would send FocusOut NotifyGrab then FocusIn NotifyUngrab. Currently
gdk would then deactivate the current surface, which makes accessibility
screen readers think that we have switched to a non-accessible application
and came back again, and thus reannounce the application frame etc. which we
don't want when e.g. just raising volume.
And actually, receiving FocusOut NotifyGrab does not mean losing the
X focus, it only means an application aqcuired a grab, i.e. it is
temporarily stealing keyboard events. On Wayland, this isn't even
notified actually.
This commit makes gdk only deactivate surfaces when there was an actual
focus switch to another window, as determined by has_focus_window (instead
of just has_focus), which happens either normally through FocusOut with
NotifyNormal, or during grabs through FocusOut with NotifyWhileGrabbed.
Fixes#85
Now that all Cairo contexts are ported to managing cairo surfaces
themselves, the old fallback code that didi the managing is no longer
needed.
Also clarify the behavior of gdk_cairo_context_cairo_create() wrt the
vfunc by doing the early exit and the clipping outside of it.
We used to pass 2 regions to GdkDrawCotnext.end_frame() but code was
confusing what they meant. So we now don't do that anymore and only pass
the region that matters: The frame region.
This makes the previous gdk_draw_context_is_drawing() function public
under a new name.
I decided against the old name because we use the term "frame" for a
drawing operation, so I wanted to have this boolean flag reuse the term.
Also, don't implement SurfaceClass.ref_cairo_surface() anymore. This
means calls to it will crash now. But as they only happen in the generic
GdkCairoContext implementation, we shouldn't be affected by that.
Plus, once all backends have been ported, that call is going away
anyway.
And make the GdkCairoContext as abstract.
The idea of this and thje following commits is to get rid of all
Cairo code in gdksurface.c (and $backend/gdksurface-$backend.c)
by moving that code into the Cairo context files.
In particular, the GdkSurfaceClass.begin_frame/end_frame()
functions (which are currently exclusively used by the Cairo code
should end up being moved to GdkDrawContextClass.begin/end_frame().
This has multiple benefits:
1. It unifies code between the different drawing contexts.
GL lives in GLContext, Vulkan in VulkanContext and Cairo in
CairoContext. In turn, this makes it way easier to reason about
what's going on in surface-specific code. Currently pretty much
all backends do things wrong when they want to sync to drawing
or to the frame clock.
2. It makes the API of GdkSurface smaller. No drawing code (apart
from creating the contexts) needs to remain.
3. It confines Cairo to the Drawcontext, thereby making it way
more obvious when backends are still using it in situations
where it may now conflict with OpenGL (like when doing the dnd
failed animation or in the APIs that I'm removing in this
branch).
4. We have 2 very different types of Cairo contexts: The X/win32
model, where we have a natively supported Cairo backend but do
double buffering ourselves and use similar surfaces and the
Wayland/Broadway model where we use image surfaces without any
Cairo backend support and have to submit the buffers manually.
By not sharing code between those 2 versions, we can make the
actual code way smaller. We also get around the need to create
1x1 image surfaces in the Wayland backend where we pretend
there's a native Cairo surface.
When asked for a nonexistent (positive) monitor number,
gdk_x11_display_get_monitor would (at best) return an uninitialized pointer,
instead of returning NULL.
That way, we can store the right region there: The actual painted area
instead of the exposed area (which is way too small).
Also, the GL context is the only user of this data, so storing it there
seems way smarter.
This ensures that the frame clock gets updated with correct presentation
times even if nothing was drawn.
This is necessary for benchmarking but would also be relevant for videos
that want to sync to the frame clock but draw frames a lot less.
