We were having a problem where property actions were
not getting state updates because prop_actions_connect
was triggered from some instance_init function while
the widget class is not in place yet.
Delay that call until the widget is fully constructed,
so we can guarantee that we are dealing with the
correct class private struct, and see all class actions.
We already dropped this invariant in gtk_widget_verify_invariants()
because it was not true in all cases. Also, it is not really useful
these days as we extended what it means to be a "child" to also
include widgets in different toplevels.
For example, a popup in a popup button need not be realized just
because the button is in a realized window. The main invariants
we want are:
* Parent is realized before child
* Widget is realized before it is mapped
This sounds like its not a huge deal, but in fact it is a massive win
for things like menus, because when we go between menus in a menubar
each switch between two open menus involves a lot of intermingled
crossing events to different surfaces and for each of these the
tooltip window of the toplevel gets assigned to the new surface. This
shouldn't be a huge deal, as the tooltip window is not even visible,
but due to the realized invariant it get re-realized each time it gets
re-assigned.
This gives us a hook to walk the widget tree whenever a global
setting changes and do per-widget invalidations. This will
replace gtk_style_context_reset_widgets().
This was only living in gtkcontainer.c for historic
reasons. Move it closer to where it belongs, and
rename it from 'idle' to 'layout', since it is
really about the layout phase of the frame clock,
nowadays.
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.
Make widgets can-focus by default, and change the semantics
of can-focus to be recursive . If it is set to FALSE, focus
can not enter the widget or its descendents at all anymore.
This commit temporarily breaks focus behavior of widgets
that did not expect to receive focus.
The only place where this should be set is when making
a widget the focus-widget of a window. We still keep
the property around in readonly form, since there are
a few places where we rely on property notification
for it.
This property doesn't carry any new information compared
to GtkWindow:focus-widget. We still keep the gtk_widget_is_focus
getter, as a convenient shortcut.
People should use shortcut controllers instead (global, capture).
A side effect of this is that GtkAccelLabel now lost its method to
magically look up accelerators to display. Somebody needs to add that
back later.
When creating shortcuts, there almost always are a trigger and an action
available for use. So make gtk_shortcut_new() take those as arguments.
Also add gtk_shortcut_new_with_arguments() so people can easily pass
those in, too.
Similar to GtkShortcutTrigger, GtkShortCutAction provides all the
different ways to activate a shortcut.
So far, these different ways are supported:
- do nothing
- Call a user-provided callback
- Call gtk_widget_activate()
- Call gtk_widget_mnemonic_activate()
- Emit an action signal
- Activate an action from the widget's action muxer
It's an outdated technology now that everybody is using GActionGroups.
If somebody wanted to support changeable shortcuts, they'd need to
reintroduce it in another way.
This adds an interface for taking care of shortcut controllers with
managed scope.
Only GtkWindow currently implements this interface, so we need to ensure
that we check if any top-level widget we reach is a shortcuts manager
before we call into it.
Allow setting the scope for a controller. The scope determines at what
point in event propagation the shortcuts will be activated.
Local scope is the usual activation, global scope means that the root
widget activates the shortcuts - ie they are activated at the very
start of event propagation (for global capture events) or the very end
(for global bubble events).
Managed scope so far is unimplemented.
This is supposed to be used to replace accelerators and mnemonics.
This is a very barebones controller that currently does nothing but
activate the binding signals. Yay.
And because we have bindings on every widget (Yes, a GtkGrid has a
keybinding - 2 in fact), we need that controller everywhere.
We need to cleanup state here immediately so that we do not potentially
access the g_class private data after it been finalized. This ensures that
the borrowed reference is dropped by the muxer.
Before this commit, adding GtkWidgetAction to class private data would
require copying the actions to each subclass as they were built or
modified. This was convenient in that it is a sort of "copy on write"
semantic.
However, due to the way that GTypeInstance works with base _init()
functions, the "g_class" pointer in GTypeInstance is updated as each
_init() function is called. That means you cannot access the subclasses
class private data, but only the parent class private data.
If instead we use a singly linked list of GtkWidgetAction, each subclass
has their own "head" yet all subclasses share the tail of the
GtkWidgetAction chain.
This creates one bit of complexity though. You need a stable way to know
which "bit" is the "enabled" bit of the action so we can track enabled
GAction state. That is easily solved by calculating the distance to the
end of the chain for a given action so that base classes sort ahead of
subclasses. Since the parent class always knows its parent's actions, the
position is stable.
A new dynamic bitarray helper also helps us avoid allocations in all the
current cases (up to 64 actions per widget) and dynamically switches to
malloc if that is to ever be exceeded.
We want access to the private data from the action muxer so we can just
move the structures to the gtkwidgetprivate.h header. Alternatively we
could create accessors, but given that we'll probably need to use this
in other areas, seems reasonable to just put it there.
People should use shortcut controllers instead (global, capture).
A side effect of this is that GtkAccelLabel now lost its method to
magically look up accelerators to display. Somebody needs to add that
back later.
Reduce the amount of special casing by using a list model
for global and managed shortcuts, too.
This way, the ListModel API will work for the ShortcutController in the
GtkShortcutManager and GtkRoot.
The only special case remaining is shortcut activation, which needs to
pass the right widget to the controller in the global/managed case.
When creating shortcuts, there almost always are a trigger and an action
available for use. So make gtk_shortcut_new() take those as arguments.
Also add gtk_shortcut_new_with_arguments() so people can easily pass
those in, too.
Similar to GtkShortcutTrigger, GtkShortCutAction provides all the
different ways to activate a shortcut.
So far, these different ways are supported:
- do nothing
- Call a user-provided callback
- Call gtk_widget_activate()
- Call gtk_widget_mnemonic_activate()
- Emit an action signal
- Activate an action from the widget's action muxer
- Activate a GAction
It's an outdated technology now that everybody is using GActionGroups.
If somebody wanted to support changeable shortcuts, they'd need to
reintroduce it in another way.
This adds an interface for taking care of shortcut controllers with
managed scope.
Only GtkWindow currently implements this interface, so we need to ensure
that we check if any top-level widget we reach is a shortcuts manager
before we call into it.
Allow setting the scope for a controller. The scope determines at what
point in event propagation the shortcuts will be activated.
Local scope is the usual activation, global scope means that the root
widget activates the shortcuts - ie they are activated at the very
start of event propagation (for global capture events) or the very end
(for global bubble events).
Managed scope so far is unimplemented.
This is supposed to be used to replace accelerators and mnemonics.
This is a very barebones controller that currently does nothing but
activate the binding signals. Yay.
And because we have bindings on every widget (Yes, a GtkGrid has a
keybinding - 2 in fact), we need that controller everywhere.
The `rename-to` annotation is used to "shadow" a symbol with another
one, which means both symbols need to exist. It can't be used to rename
a symbol to something else.
This is a huge reorganization of GtkDropTarget. I did not know how to
split this up, so it's unfortunately all one commit.
Highlights:
- Split GtkDropTarget into GtkDropTarget and GtkDropTargetAsync
GtkDropTarget is the simple one that only works with GTypes and offers
a synchronous interface.
GtkDropTargetAsync retains the full old functionality and allows
handling mime types.
- Drop events are handled differently
Instead of picking a single drop target and sending all DND events to
it, every event is sent to every drop target. The first one to handle
the event gets to call gdk_drop_status(), further handlers do not
interact with the GdkDrop.
Of course, for the ultimate GDK_DROP_STARTING event, only the first
one to accept the drop gets to handle it.
This allows stacking DND event controllers that aren't necessarily
interested in handling the event or that might decide later to drop
it.
- Port all widgets to either of those
Both have a somewhat changed API due to the new event handling.
For the ones who should use the sync version, lots of cleanup was
involved to operate on a sync API.
And in particular, only do it if the widget doesn't use ALIGN_FILL.
This avoids lots of measuring in the common case and speeds up
size_allocate() by about 25%.
And because size_allocate() is the bottleneck in the fishbowl, this also
gets ~25% more fishies.
It's the native's job to request a 1px x 1px size, not the job of
gtk_widget_size_allocate()
Also saves 10% of size_allocate() time because checking for an interface
is really expensive.
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.
Instead of relying on gdk's antiquated crossing events,
create a new GtkCrossingData struct that contains the
actual widgets, and a new event controller vfunc that
expects this struct. This also saves us from making sense
of X's crossing modes and details, and makes for a
generally simpler api.
The ::focus-in and ::focus-out signals of GtkEventControllerKey
have been replaced by a single ::focus-change signal that
takes GtkCrossingData as an argument. All callers have
been updated.
We want to make events readonly, so stop translating
their coordinates and instead pass the translated
coordinates separately, when propagating events.
In Vala and JS at least, gtk_widget_get_css_name() and
gtk_widget_class_get_css_name() are resolved to
GtkWidget.get_css_name().
To avoid this problem, we rename the class version.
These don't take a duration, instead they call g_get_monotonic_time() to
and subtract the start time for it.
Almost all our calls are like this, and this makes the callsites clearer
and avoids inlining the clock call into the call site.
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.
usec is the scale of the monotonic timer which is where we get almost
all the times from. The only actual source of nsec is the opengl
GPU time (but who knows what the actual resulution of that is).
Changing this to usec allows us to get rid of " * 1000" in a *lot* of
places all over the codebase, which are ugly and confusing.
Hide the expensive GTK_IS_ROOT() checks behind the cheaper realized,
mapped, etc. checks. This way we only check for the widget being a root
if the invariant does not hold.
This gets pop_verify_invariants() from 16% to 6% when running the
scrolling benchmark in the widget-factory.
1. Rename the thing
2. Turn it from a signal to a vfunc
3. Pass the GtkCssStyleChange to it
We don't export any public API about the GtkCssStyleChange yet, it's
just a boring opaque struct.
Replace uses of gtk_css_style_get_value with direct access,
throughout the tree. We don't replace all uses, just those
where we are dealing with a fixed property. Be careful to
handle the currentColor special case for color properties.
Add GtkWidget API for adding and removing style classes, as well as
checking whether a widget has a style class applied.
Everyone has to go through GtkStyleContext for this these days but with
GtkStyleContext eventually going away, it makse sense for GtkWidget to
have API for this.
The reason for this is simply that I want to get hash functions that
have their values close together, so they can fit in a smaller range
(the goal here is 12 bits). By using GQuark, we get consecutive numbers
starting with 1 (and applications have <1000 quarks usually), whereas
interned strings can be all over the place.
As a side effect we also save 64 bytes per declaration.
Previously, we wrapped all GtkCssShadowValues in a GtkCssShadowsValue,
even if it was just one shadow. This causes an unnecessary bloat in
css values.
Make each GtkCssShadowValue able to handle multiple shadows instead, and
use gtk_css_shadow_value* API everywhere.
I was seeing crashes in gtk_widget_run_controllers.
We were accessing the controller after calling out
to application code that might remove it. Better
be safe and do the access before.
GtkBuilderScope is an interface that provides the scope that a builder
instance operates in.
It creates closures and resolves types. Language bindings are meant to
use this interface to customize the behavior of builder files, in
particular when instantiating templates.
A default implementation for C is provided via GtkBuilderCScope (to keep
with the awkward naming that glib uses for closures). It is derivable on
purpose so that languages or extensions that extend C can use it.
The reftest code in fact does derive GtkBuilderCScope for its own scope
implementation that implements looking up symbols in modules.
gtk-widget-factory was updated to use the new GtkBuilderCScope to add
its custom callback symbols.
So it does it different from gtk-demo, which uses the normal way of
exporting symbols for dlsym() and thereby makes the 2 demos test the 2
ways GtkBuilder uses for looking up symbols.
Use it as the default object for expression binds and when connecting
signals. It is intended to work kind of as the "this" object while
parsing. In fact, the term "current object" was stolen from the Java
docs and various C++ tutorials for the this pointer.
Set the current object in gtk_widget_init_template() and
GtkListItemBuilder.
This more-or-less replaces the object passed to
gtk_builder_connect_signals() in GTK3.
gtk_builder_connect_signals() is no longer necessary, because all the
setup that made it necessary to have this extra step is now done
automatically via the closure functions.
This is pretty unused and gets in the way of the next steps.
A potential side effect is that for templates the widget was passed as
the user data argument. If that turns out to be important, we have to
special case that situation.
... inside the default vfunc.
Instead, walk the children until we find the first widget that can be
focused. If no child can be focused, return FALSE from grab_focus.
The value returned by gtk_widget_get_settings() depends on the widget's
display, so watch for notify::display instead of using (un)root for
this.
Fixes the warnings seen when show a file chooser from the inspector.
I recently turned gtk_widget_activate_action()
into a varargs function. That is more convenient
from C, but we need a non-varargs variant for
bindings. So add the old API back, under the
name gtk_widget_activate_action_variant(),
with a rename-to annotation.
We need to create a muxer eagerly for every
widget that has class actions, since those
are otherwise missed in the action lookup
on the muxer side. But otherwise, there is
no reason to create parent muxers aggressively,
as long as we update the parent muxers on
root/unroot.
This reduces the number of muxers we create
in widget-factory from 210 to around 50.
The only cases of stateful actions we've seen
so far have been boolean properties, and we
don't really want to add much state handling
API, so lets just go with property actions
for now.
Adapt the only user in GtkText.
Add a facility to register and install actions
at class init time. The intended use for these
actions is for
a) context and other model-based menus
b) key bindings
Most of these actions are going to be stateless,
so add separate apis for the simple and stateful
cases.
We avoid creating an action group for these by
teaching the action muxer about these actions.
The action muxer also maintains the enabled
state for these actions.
Add a convenience api to skip children
that should not be included in the layout,
and call gtk_native_check_resize on all
native children outside of the vfunc.
Make gtk_widget_real_focus do the right
thing for focusable widgets with children.
A case where this is (now) relevant is
an entry with a context popover.
We don't need to cover every case with a va_marshaller, but there are a
number of them that are useful because they will often only be connected
to by a single signal handler.
Generally speaking, if I opened into a file to add a va_marshaller, I just
set all of them.
Similar to previous removals of g_cclosure_marshal_VOID__VOID we can remove
other marshallers for which are a simple G_TYPE_NONE with single parameter.
In those cases, GLib will setup both a c_marshaller and va_marshaller for
us. Before this commit, we would not get a va_marshaller because the
c_marshaller is set.
Related to GNOME/Initiatives#10
The CSS transform should operate on the border-box, not the margin box.
So we need to shrink the bounds by the margin before we apply the CSS
transform.
Make the transform (transfer full).
1. This makes sure we actually reference the transform. Previously we
did not.
2. Most callers create a new transform to pass to us. Now they don't
have to uref it anymore.
When a GtkNative widget is marked as resize_needed,
we need a current position for its parent and we need
the parent to be allocated (so we can position our
surface), but we don't need the parent to be marked
as resize_needed, since the parent size is entirely
independent of the popup size.
When we print warnings about a widget, using
gtk_widget_get_name() is slightly better than
G_OBJECT_TYPE_NAME(), since it will give us
the widgets unique name when available.
Now that roots can have parent widgets, we need to
carefully examine all calls of gtk_widget_get_toplevel,
and replace them with gtk_widget_get_root if we want
the nearest root, and not the ultimate end of the parent
chain.
