GitLab

This is in preparation to modernizing our handing
of exported symbols.

glib-mkenums is not currently clever enough to know which version an
enum type was added in, so just mark all the _get_type() functions as
available in all versions.

Add a macro to declare that a particular symbol is available in all
versions of GTK+.

All newly-added symbols should have proper version macros (like
GDK_AVAILABLE_IN_3_4).

If config.h is included after gdkversionmacros.h, _GDK_EXTERN
gets redefined and cpp doesn't like that.

It is not needed here.

_gtk_text_layout_get_line_xrange was not exported and not used
inside GTK+ either.

This avoids an evil trap when doing MAX (..., ... - 2 * border_width)
and the expression on the right gets promoted to unsigned, instead
of going negative as you would expect.
https://bugzilla.gnome.org/show_bug.cgi?id=699633

Draw a checkerboard pattern, and add an interactive mode that
waits for events before moving on. This makes it easier to see
when a size is wrong.

These tests check that a toplevel window ends up with the expected
size after setting default sizes or resizing. It currently passes
on X, but fails with client-side decorations under X and Wayland.

When we call _gdk_wayland_display_load_cursor_theme during
the initial opening of the first display, gdk_setting_get does
not work yet, since it relies on the default display/screen
being set, which only happens after open returns.
Instead, just use the screen of this display.

This was broken in e9f182e3

This makes GtkSettings survive the object finalize test.

Document that the restrictions on Win32 apply also to Quartz.

Otherwise Emacs makes one's life miserable.

GtkIconView doesn't ref the rows of the model so we have to make sure that a
level is actually empty before destroying it.

https://bugzilla.gnome.org/show_bug.cgi?id=698846

This was broken since forever. Good that we never finalize
theme engines...

As far as refcounting is concerned, invisibles are like
windows: you have to destroy them.

This makes the file chooser pass this test

If the async call is finished after the dialog is already disposed,
bad things happen. Prevent that by marking the dialog as dismissed
in this case.

It's not necessary anymore because gdk_display_manager_get() always
succeeds and the value is independant of when it was called as it's no
longer backend specific.

Fixes the window not emitting a "notify::attached-to" signal when the
attach-widget goes away.

This is not needed anymore, as only one type exists and that type can
always be instantiated.

This is an API break, but the broadway backend is unsupported, so we can
get away with it.

This is an API break, but the wayland backend is unsupported, so we can
get away with it.

Just create a backend-agnostic displaymanager. Creating the displays
later on determines the backend in use.

Use the same logic for display creation in
gdk_display_manager_open_display() that is used for displaymanager
creation.

This follows the same reasoning as the X11 backend in commit
0122a9da

This follows the same reasoning as the X11 backend in commit
0122a9da

AtkSelection requires that the immediate children of the object are the
selectable items. The combobox however is implemented with just 1 child:
The popup menu.

The popup menu is implementing the selectable interface.

Test are updated to reflect this change.

When the hostnamed call is cancelled, we get a G_IO_ERROR_CANCELLED
error back. Handle it properly.

This reverts commit f326c0ea.

Grr, turns out that cancelling is not that reliable :-(

We were silently relying on everybody to detach their windows,
and would leak memory otherwise.

Taking a full reference of the attach widget creates a reference
cycle, and keeps the attach_widget alive longer than it should.

If we cancel the call in dispose, no need to ref the object
we pass as userdata. Avoids the risk of leaking the ref if
the call is actually canceled.

This reverts commit cf669508.

This reverts commit 17b97333.