We don't want to emit state changes for all the cells in a row, just for
the cell in the expander column. It's the only one that reports EXPANDED
or EXPANDABLE states, after all.
Also, contains refactoring of the affected functions for all the special
cases.
https://bugzilla.gnome.org/show_bug.cgi?id=694059
We weren't checking the removed flag but the added flag for removal
emissions, so what would happen for every state change notification was:
- on state-added, both an "added" and a "removed" event were emitted
- on state-removed, nothing
https://bugzilla.gnome.org/show_bug.cgi?id=694059
The callback function gtk_window_on_theme_variant_changed is only used on the
X11 backend (where GtkSettings is used for the settings information.)
Fixes: https://bugzilla.gnome.org/show_bug.cgi?id=674207
We can't safely examine allocations synchronously using
gtk_main_iteration(), as there might be not enough time for a new paint
clock tick to have expired and the allocation set on the widget.
Work this around adding g_usleep() calls before processing pending
mainloop events.
With the following code:
#define INVALID_CHAR GDK_KEY_VoidSymbol - 1
gtk_accelerator_get_label (INVALID_CHAR, GDK_SHIFT_MASK | GDK_CONTROL_MASK);
we would get this label:
Shift+Ctrl+
instead of this label:
Shift+Ctrl
https://bugzilla.gnome.org/show_bug.cgi?id=694075
Add an API to start or stop continually updating the frame clock.
This is a slight convenience for applcations and avoids the problem
of getting one more frame run after an animation stops, but the
primary motivation for this is because it looks like we might have
to use timeBeginPeriod()/timeEndPeriod() on Windows to get reasonably
accurate timing, and for that we'll need to know if there is an
animation running.
https://bugzilla.gnome.org/show_bug.cgi?id=693934
We only draw the main entry on should_draw (widget->window), because
otherwise we also draw it on the GtkTextHandle widgets.
This is necessary due to the recent change for that to not return
TRUE and swallow the rest of the drawing operation.
This was causing warnings on widget unparent like:
Gdk-CRITICAL **: gdk_window_has_native: assertion `GDK_IS_WINDOW (window)' failed
Becasue the window was not properly removed from the lists on unrealize.
We clear GtkTickCallbackInfo on creation to ensure all fields start
as 0. Before we sometimes ended up with destroyed being 1
so the tick was never called.
gtk_icon_info_copy and gtk_icon_info_free are deprecated for
the corresponding GObject methods.
We set correct transfer markup for the GtkIconInfo returning methods
to fix the introspection of them.
gtk_icon_info_load_symbolic_for_context_async had the wrong method
name in its documentation block.
We need to disconnect the frame clock when we unrealize (at which
point the old clock is still alive) not in destroy(). Since there
is no common unrealize for containers, trigger this from GtkWidget.
Add a very simple GtkWidget function for an "tick" callback, which
is connected to the ::update signal of GdkFrameClock.
Remove:
- GtkTimeline. The consensus is that it is too complex.
- GdkPaintClockTarget. In the rare cases where tick callbacks
aren't sufficient, it's possible to track the
paint clock with ::realize/::unrealize/::hierarchy-changed.
GtkTimeline is kept using ::update directly to allow using a GtkTimeline
with a paint clock but no widget.
Deprecate gdk_window_enable_synchronized_configure() and
gdk_window_configure_done() and make them no-ops. Implement the
handling of _NET_WM_SYNC_REQUEST in terms of the frame cycle -
we know that all processing will be finished in the next frame
cycle after the ConfigureNotify is received.
When we have a looping animation for something like an angle,
we need to make sure that the distance we go past 1.0 becomes
the starting distance for the next frame. This prevents a
stutter at the loop position.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Use GdkFrameClock for the timing of GtkTimeline. This require the
user to provide either a GtkWidget or a GdkFrameClock when creating
the timeline. The default constructor now takes a GtkWidget. If you
want to create a GdkFrameClock without a widget, you need to use
g_object_new() and pass in a GdkFrameClock and GdkScreen.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
The frames-per-second for an animation should be controlled by how
fast we can process frames and the the frame-rate of the display; it's not
a meaningful app-settable property.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Add back the GtkTimeline code that previously made private and
then removed. It will be hooked up to GdkFrameClock. This commit
purely adds the old code back.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Switch GtkStyleContext to using GdkFrameClock. To do this, add a new
UPDATE phase to GdkFrameClock.
Add a GdkFrameClockTarget interface with a single set_clock() method,
and use this to deal with the fact that GtkWidget only has a frame
clock when realized.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
Instead of having gdk_frame_clock_request_frame() have
gdk_frame_clock_request_phase() where we can say what phase we need.
This allows us to know if we get a frame-request during layout whether
it's just a request for drawing from the layout, or whether another
layout phase is needed.
https://bugzilla.gnome.org/show_bug.cgi?id=685460
This is necessary in order to have async operations on it.
All the old copy/free functions keeps working, and g_boxed_copy on a GObject
also works, so this should be mostly compatible, but techncally its a minor
ABI break since the GType changes fundamental type. Changes like this has
happened before though, like with GVariant becomming its own fundamental
type.
https://bugzilla.gnome.org/show_bug.cgi?id=693802
These are generic tests that can test the button in all of its modes,
instead of hand-written tests for each combination.
Some tests fail currently.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
If the user didn't explicitly select anything, BUT the file chooser button has
a current_folder set, do the same as what GtkFileChooserDefault would do:
return the current folder as the selection.
This makes the tests in tests/filechooser pass!
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
If no file was originally selected in the GtkFileChooserButton, then its
internal dialog is brought up and cancelled, then we need to restore the
selection back to none. GtkFileChooser, though, doesn't like to
select a NULL file, so call _unselect_all() in that condition.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
The button's underlying file chooser dialog should not be used to store the file selection
while the dialog is unmapped. Instead, the file chooser button now stores the
selection itself.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
It used to fetch a possibly multiple selection from the GtkFileChooserDialog, and then
pick just the first item from the selection list. But since GtkFileChooserButton
operates in single-selection mode only, it can simply use gtk_file_chooser_get_file()
instead.
Also, the right way to reset the selection for GTK_FILE_CHOOSER_ACTION_SELECT_FOLDER
is with gtk_file_chooser_select_file(), not with _set_current_folder_file().
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This code came from a home-grown testing mechanism, which didn't aggregate tests
into a test suite; it just ran them one by one. Here we move some of that machinery
to GTestDataFunc for more flexibility in running tests.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This adds a way to get the gtk_widget_set_opacity liike behaviour
of retargeting GdkWindows and exposing every child in ::draw, without
actually having an alpha. This is needed if you're doing more complex things
such as cross fading of widgets.
We do this as a hack by using opacity values that round to 255 yet not
really 1.0 in order to avoid having some magical API call for this
mainly internal call.
https://bugzilla.gnome.org/show_bug.cgi?id=687842
Sink the GtkEntry assigned to the private structure of GtkCellRendererText
before signals containing it as an argument are sent out. This keeps
language bindings from sinking the reference and then destroying the entry
when the signal closure is finished.
https://bugzilla.gnome.org/show_bug.cgi?id=693400
When calling gtk_widget_draw() on the entry gtk_cairo_should_draw_window()
will return TRUE for all windows. This is used when rendering a widget to
somewhere other than the screen, and its now used for transparent widgets.
This caused the texthandle to always draw itself and terminate the draw
handler for the entry.
Instead we now only draw the markers when really visible, plus we return
FALSE to avoid stopping the entry drawing.
https://bugzilla.gnome.org/show_bug.cgi?id=687842
This adds gtk_widget_get/set_opacity, as well as a GtkWidget.opacity
property. Additionally it deprectates gtk_window_get/set_opacity and
removes the GtkWindow.opacity property (in preference for the new
identical inherited property from GtkWidget, which should be ABI/API
compat).
The implementation is using the new gdk_window_set_opacity child
window support for windowed widgets, and cairo_push/pop_group()
bracketing in gtk_widget_draw() for non-window widgets.
https://bugzilla.gnome.org/show_bug.cgi?id=687842
This replaces the previously hardcoded calls to gdk_window_set_user_data,
and also lets us track which windows are a part of a widget. Old code
should continue working as is, but new features that require the
windows may not work perfectly.
We need this for the transparent widget support to work, as we need
to specially mark the windows of child widgets.
https://bugzilla.gnome.org/show_bug.cgi?id=687842
The "activate" action here did not do anything.
It is possible we actually want to have some actions here,
like "step-up", "step-down", "page-up", "page-down", etc.
For now, just remove the AtkAction implementation.
https://bugzilla.gnome.org/show_bug.cgi?id=553334
There's really no reason why we shouldn't automatically create a
GtkViewport when the widget added to GtkScrolledWindow is not a
GtkScrollable, instead of just printing a g_warning.
Copy the viewport special case into the scrolled window implementation
of gtk_container_add().
https://bugzilla.gnome.org/show_bug.cgi?id=693015
Saves ~6MB of memory per application in the Adwaita I am using - at
least until the app starts using all the images in the theme, because
the code doesn't discard images yet once they were loaded.
https://bugzilla.gnome.org/show_bug.cgi?id=692934
This is essentially a GtkCssImage for a cairo_surface_t and is a pretty
much straight up copy of GtkCssImageUrl. But we want to implement lazy
loading and animations, so GtkCssImageUrl is going to gain new
features...
https://bugzilla.gnome.org/show_bug.cgi?id=692934
I'd like to use it when printing the value, but I haven't found a way to
do that sanely yet, as I'd need to be able to print relative paths for
make check to work (otherwise the srcdir would blow things up). And we
use a GString to output to, so there's no way to attach a base dir to
that.
If anyone has an idea how to achieve that, poke me. Having the real
filename in debug prints sounds like a very good idea to me.
https://bugzilla.gnome.org/show_bug.cgi?id=692934
Commit ddceddaa84 removed the call to
gtk_style_context_set_background() in favour of always rendering it with
gtk_render_background() during the draw vfunc.
This has the side effect of making the backing window always
transparent, which blocks GTK from applying some optimizations during
the paint cycle. The result is that, especially in clutter-gtk
applications, scrolling performance gets really bad.
This commit partially reverts ddceddaa84
and changes the code so that both gtk_style_context_set_background() and
gtk_render_background() are called
Commit da09447914 removed the call to
gtk_style_context_set_background() in favour of always rendering it with
gtk_render_background() during the draw vfunc.
This has the side effect of making the backing window always
transparent, which blocks GTK from applying some optimizations during
the paint cycle. The result is that, especially in clutter-gtk
applications, scrolling performance gets really bad.
This commit partially reverts da09447914
and changes the code so that both gtk_style_context_set_background() and
gtk_render_background() are called.
gtk_tree_drag_source_drag_data_get's GtkSelectionData argument should not be
marked as (out) because:
a) GtkSelectionData is semi-private (it's declared in gtkselectionprivate.h),
and thus gobject-introspection has no knowledge of its fields or its size.
There is thus no way for language bindings to allocate GtkSelectionData.
b) Even if it was possible for language bindings to allocate GtkSelectionData,
a zeroed-out instance thus created would not be usable with
gtk_tree_drag_source_drag_data_get. As far as I can tell, you need to
initialize its "target" member to the GdkAtom of "GTK_TREE_MODEL_ROW".
Language bindings have no way of knowing this, of course.
https://bugzilla.gnome.org/show_bug.cgi?id=692844
This is akin to commit cfb09e5654 in the gtk-2-24 branch;
the last_folder_uri is no longer being used for anything meaningful, so we
remove it altogether.
Signed-off-by: Federico Mena Quintero <federico@gnome.org>
This is a quickfix to keep things working.
It turns out GtkWindow assumes it can do sizing operations while not
being visible, or while in the process of show()ing/hide()ing itself.
And commit b495ce54 broke these operations.
Figuring this properly requires some more thinking and restructuring on
my part, so for now we relax the requirement of visiblility enough for
these things to start working again.
The code is always instantiating this schema at a fixed location, so why
is it relocatable?
Add a path so that it shows up properly in dconf-editor, and from the
gsettings commandline tool.
The code is always instantiating this schema at a fixed location, so why
is it relocatable?
Add a path so that it shows up properly in dconf-editor, and from the
gsettings commandline tool.
https://bugzilla.gnome.org/show_bug.cgi?id=692163
-Rename the "libgail" projects to gtka11y, for consistency with the
autotools builds
-Update the projects completion in gtk/a11y/Makefile.am, as the sources are
now listed under $(libgtka11y_la_SOURCES) instead of $(libgail_la_SOURCES)
An instance of GtkAdjustment may be used by another instance after
the spin button widget is destroyed. In that case, the function
gtk_spin_button_accessible_value_changed() will be called with an
invalid argument. This situation is often caused when one use
GtkCellRendererSpin widget. To avoid invalid call of the function,
the signal handler for the "value-changed" signal should be disconnected
when the spin-button widget is destroyed.
Using g_signal_connect_object achieves just that.
https://bugzilla.gnome.org/show_bug.cgi?id=691592
Since 16195ad the “expand” property is always set to FALSE when a
column is resized. This commit takes a different approach and enables
“expand” whenever the column is wide enough. An appropriate
“fixed-width” (so that the desired width is achieved after expanding) is
calculated using equations that are explained in the code.
https://bugzilla.gnome.org/show_bug.cgi?id=691751
Rewrites gtk_tree_view_column_request_width() and
gtk_tree_view_size_allocate_columns() to respect the minimum and natural
sizes that are already being returned by
gtk_cell_area_context_get_preferred_width().
The convoluted logic explained (not!) by this comment has been removed:
“Only update the expand value if the width of the widget has changed, or
the number of expand columns has changed, or if there are no expand
columns, or if we didn't have an size-allocation yet after the last
validated node.” This logic seems to have been a workaround for the
“jumping” behavior fixed in 16195ad and is no longer necessary.
https://bugzilla.gnome.org/show_bug.cgi?id=691751
Removes the hidden “resized-width” and “use-resized-width” properties
from GtkTreeViewColumn and instead uses the “fixed-width” property to
serve the same purpose. “fixed-width”, if set, will now override the
auto-sized width (-1 is now a legal value meaning “not set”).
Additional “cleanups” in this commit:
1. When the user resizes the column the “expand” property is now also
set to FALSE, in order to prevent the column from suddenly jumping to a
different width when the window is resized.
2. The code that translated mouse movement to column sizes has been
simplified:
the change in column width is now calculated directly from the distance
the mouse cursor has traveled. Weird behavior that might have happened
previously if the position of the column changed during resizing, is now
prevented.
3. There was some lengthy logic handling the keyboard shortcuts used to
resize treeview columns, which would call gtk_widget_error_bell() once
the minimum or maximum width was reached. Instead of rewriting these
checks I simply set the “fixed-width” property to what was requested,
relying on the fact that it is already clamped between the minimum and
maximum width during size allocation.
I will greatly surprised if anyone notices the missing error bell.
https://bugzilla.gnome.org/show_bug.cgi?id=691751
Splits up size_request() so that the height calculations are only done
when get_preferred_height() is called and the width calculations are
only done when get_preferred_width() is called. Since
get_preferred_width() does not change the treeview->priv->width value,
treeview->priv->prev_width will always be equal to it and can therefore
be removed. The only place where prev_width was used is a block in
gtk_tree_view_size_allocate(). This block seems to be adjusting the
horizontal scrollbar to account for treeview->priv->width having been
changed in size_request() and should no longer be necessary. A similar
block immediately above it seems to already account for the width change
in size_allocate().
https://bugzilla.gnome.org/show_bug.cgi?id=691751
After “validation” (i.e., background size calculations) of some cells,
size_request() was called here to update the internally cached size of
the treeview. Apparently not updating the sizes leads to some kind of
“inconsistency” that messes with top_row_to_dy(). In the GTK3 model for
size allocation, things are more complicated. The treeview can’t just
go ahead and calculate its own size any more; instead it reports both a
“minimum” and a “natural” size, and it doesn’t know what size it will
actually get until size_allocate(). It may be necessary to update
top_row_to_dy() to deal with not knowing the exact size.
https://bugzilla.gnome.org/show_bug.cgi?id=691751
We want to reserve space for the size of the scrollbars even when they
are not visible. And because toggling visibile to off now returns 0 for
size requests, this won't work anymore.
The window size can be queried on widget->window directly, no need to
store it in widget->allocation.
This change is necessary because gtk_widget_set_allcation() is now
checking invariants that assume it's called from insize
gtk_widget_size_allocate() and that wasn;t the case here.
Commit e32da246a8 made GtkRange's trough
respect the CSS margin property, but it also trimmed the box in which
the trough reacts to click events by the margin.
We still want to catch events in that area instead, and just make sure
the margin is applied when drawing (which was already implemented by
that commit).
This commit reverts the parts of
e32da246a8 that didn't involve drawing,
fixing the bug.
https://bugzilla.gnome.org/show_bug.cgi?id=691677
When cross-compiling, instead of depending on a natively built GTK+ (which means
building Glib, ATK, Pango, gdk-pixbuf, libX11...) for gtk-update-icon-cache,
find the host compiler and gdk-pixbuf, and build another gtk-update-icon-cache
with that.
This uses AX_PROG_CC_FOR_BUILD from autostars to find the host compiler, and
assumes that you'd set PKG_CONFIG_FOR_BUILD to a host pkg-config binary.
https://bugzilla.gnome.org/show_bug.cgi?id=691301
GtkWidget::visible is required for the widget to:
- have a preferred size other than 0/0
- have a size allocated
- return other values than { -1, -1, 1, 1 } from get_allocation()
This is an experimental patch aiming to make concepts and behaviors
inside GTK more concreate. GtkWidget::visible is now essentially what
CSS does for "display: none".
Note that if you want the effect of CSS's "visibility: hidden", you'll
have to use a GtkNotebook with an empty page as the concept of reserving
space but not drawing anything isn't supported natively in GTK.
It's a lot uglier now, but it shouldn't crash anymore.
We must update the font description for animations, but we can't free it
on query, because some paths call gtk_style_context_get_font() twice in
a row without stopping the use of the first call. So us just creating a
new font description all the time and unreffing the old one is not a
good idea. So we just mere the new one into the old one.
https://bugzilla.gnome.org/show_bug.cgi?id=691186
Previously, with STATE_FLAGS_REPLACE we would unset _all_ the state
flags on children, not just the ones that do propagate. This caused the
RTL/LTR flags to get lost.
We add a separate gtk-a11y.h single-include header for
them. This header will work much the same as gtkx.h. It
will be installed in /usr/include/gtk-3.0/gtk, but you
have to include it separately.
Since we are going to install these headers soon, we need
to make their mutual includes work in the installed location
as well. Also, avoid including individual gtk headers, to
avoid trouble with single-include guards.
This commit exposes the get_type() functions and standard
headers for accessible implementations. This makes it possible
to derive from the GTK accessible implementations without
GType magic tricks. This is necessary, because we require the
a11y type hierarchy to be parallel to the widget type hierarchy.
So, if you derive a widget and need to adjust its a11y implementation,
you have to be able to derive its accessible implementation.
This commit probably exposes more than is absolutely necessary,
it also exposes accessibles of widgets that are unlikely candidates
for deriving from.
Since not every theme renders a background for a GtkViewport (and
Adwaita master doesn't), ensure the grid+viewport we use to emulate a
text view here uses the "view" style class.
It already paints the css border, so let's make it also honor css
background. This is needed to have a box of a different color around
some widgets (e.g. latest gnome-clocks design)
If you want to get rounded corners on an hbox, instead of
:first-child {
border-top-left-radius: 5px;
border-bottom-left-radius: 5px;
}
:last-child {
border-top-right-radius: 5px;
border-bottom-right-radius: 5px;
}
you now need to write:
:first-child, :last-child:dir(rtl) {
border-top-left-radius: 5px;
border-bottom-left-radius: 5px;
}
:last-child, :first-child:dir(rtl)
{
border-top-right-radius: 5px;
border-bottom-right-radius: 5px;
}
If you want to get rounded corners on an hbox, instead of
:first-child {
border-top-left-radius: 5px;
border-bottom-left-radius: 5px;
}
:last-child {
border-top-right-radius: 5px;
border-bottom-right-radius: 5px;
}
you now need to write:
:first-child, :last-child:dir(rtl) {
border-top-left-radius: 5px;
border-bottom-left-radius: 5px;
}
:last-child, :first-child:dir(rtl)
{
border-top-right-radius: 5px;
border-bottom-right-radius: 5px;
}
If you want to get rounded corners on an hbox, instead of
:first-child {
border-top-left-radius: 5px;
border-bottom-left-radius: 5px;
}
:last-child {
border-top-right-radius: 5px;
border-bottom-right-radius: 5px;
}
you now need to write:
:first-child, :last-child:dir(rtl) {
border-top-left-radius: 5px;
border-bottom-left-radius: 5px;
}
:last-child, :first-child:dir(rtl) {
border-top-right-radius: 5px;
border-bottom-right-radius: 5px;
}
This function is just a sophisitcated optimization.
If we know the GDK window's background will be opaque, we mark it as
opaque. This is so GDK can do all the optimizations it does for opaque
windows and be fast.
This is mainly used when scrolling.
The previous code didn't get this right, in particular it didn't enforce
a transparent background when it knew the background was not opaque.
The code path where we update the tooltip text property doesn't set
the state and value variables, and so doesn't need to call
notify_state_change().
Return early, and move the if block at the beginning of the function for
clarity.
There are some registred stock ids like gtk-discards that have no icons,
and you could also pass a non-registred stock id. Both of these means
gtk_style_context_lookup_icon_set returns NULL, which causes
a critical in gtk_icon_set_render_icon_pixbuf.
We avoid this by just making these render as EMPTY.
We used to use GTK_RESIZE_QUEUE, but that is problematic for e.g
a GtkScrolledWindow with NEVER scroll policies, as size changes
in ancestors will never get propagated to the scrolled window, causing
it to not have the correct size.
This is a slight performance hit, but in practice its not bound to be
problematic. In typical UIs there is only a single "large" GtkScrolledWindow
visible at a time, so a size requeust propagating out of such a window
will only hit the smaller amount of widgetry outside the scrolled window,
and additionally all such widgets will have their size request caches
still valid.
https://bugzilla.gnome.org/show_bug.cgi?id=690099
We don't get an automatic queue resize on realize anymore, which
was papering over this bug where we did not set the child window
size/position at realize time.
We use the new g_type_get_type_registration_serial() so that we can
cache and properly invalidate the result of g_type_from_name().
This bumps the glib requirement to 2.35.3 to get the new function.
https://bugzilla.gnome.org/show_bug.cgi?id=689847
Rather than first collecting matches and then getting the change
for them we do the change collection directly on the tree. This
is about twice as fast.
At the moment, gtk+ doesn't depend on intltool, which is the program
that knows how to translate schemas. Attempting to translate them
causes a build failure, so for now, let's leave them in en_US.
https://bugzilla.gnome.org/show_bug.cgi?id=689584
We must not release the GtkClipboardOwner in pasteboardChangedOwner
becaue we don't own a reference to ourselves (NSPasteboard does).
Instead, release the owner right after setting it, transferring
ownership to NSPasteboard
Also, fix repeated setting of the same owner by keeping the
owner around in GtkCLipboard, and re-use it if "user_data"
doesn't change. To avoid clipboard_unset()ting our own contents
in the process, add an ugly "setting_same_owner" boolean to
GtkClipboardOwner, set it during re-setting the same owner,
and avoid calling clipboard_unset() from pasteboardChangedOwner
if it's TRUE.
(cherry picked from commit 4a8df7a33c)
We currently invalidate the whole tree every time the style state
changes in the tree view. The primary reason for this is to catch
default font changes as that may affect text cell renderers. But
cell renderers could *potentially* also read other style properties
(although that seems weird and unlikely).
We handle this by invalidating only when some state that affects sizes
is changed. This includes all the font properties.
With pango handling changes to the PangoLayout there now is no
style changes that can affect the layout for the entry, so we don't
have to reset the layout whenever the style is updated.
Now that Pango tracks changes to the context automatically there is
no need to do it manually in e.g. style-updated or direction-changed,
in fact the only case we have to care about is when we re-create
the PangoContext due to a screen change, so we only have to clear
the layouts in GtkLabel in screen-changed.
This means we're not clearing all the layouts whenever the state changes,
which happens to every widget when the window is unfocused, which helps
performance a lot.
https://bugzilla.gnome.org/show_bug.cgi?id=340066
This is for a very simple reason: The getter is returning a const value
and the font isn't const anymore. So we need to store the font
description somewhere but we can't reuse it as it's changing all the
time (yay animations, yay inherited values). Sucks.
So keep the hack in here but deprecate the function.
Instead of using gtk_style_context_get_font() in
pango_context_get_metrics(), use pango_context_get_font_description().
The context contains the font description we are about to use after all.
