The mechanism of setting label padding had to be changed consequently.
It fixes following warnings:
(gnome-control-center:32638): Gtk-WARNING **: GtkComboBox 0x28bcef0 is mapped but not child_visible
(gnome-control-center:32638): Gtk-WARNING **: GtkComboBoy 0x28bcef0 is mapped but visible=1 child_visible=0 parent GtkNotebook 0x2b86600 mapped=1
The ugly hack to catch when combo is losing focus had to be also removed
and notify::popup-shown is used instead. It fixes when GtkComboBox is wrongly
shown instead of GtkButton due to escape is pressed or active item is selected.
https://bugzilla.gnome.org/show_bug.cgi?id=694325
The mechanism of setting label padding had to be changed consequently.
It fixes also following warnings:
(gnome-control-center:32638): Gtk-WARNING **: GtkEntry 0x28bcef0 is mapped but not child_visible
(gnome-control-center:32638): Gtk-WARNING **: GtkEntry 0x28bcef0 is mapped but visible=1 child_visible=0 parent GtkNotebook 0x2b86600 mapped=1
https://bugzilla.gnome.org/show_bug.cgi?id=694325
... so that it is possible to override virtual functions and change
the order of the object creation and destruction process without having
to write a bunch of prototypes. Having the class_init below the others
helps us do that.
https://bugzilla.gnome.org/show_bug.cgi?id=731713
When rotating the left-most monitor, we could end up with one
of two scenarios:
- Huge gap between monitor one and two when the left monitor is made
to be vertical
- Monitor two overlapping on top of monitor one when monitor is
back to being horizontal.
In addition to adding a bit more debug, we now take the old and new
rotation of monitor one into account when calculating the moving
screen edges.
https://bugzilla.gnome.org/show_bug.cgi?id=731166
Consider whether rotation has changed when checking whether
we need to relayout the screens. Before, we would only have changed the
layout if the resolution changed.
This doesn't fix the problem of the relayouting code not taking rotation
into account, but it will at least allow us to run that code when only
the rotation changes.
https://bugzilla.gnome.org/show_bug.cgi?id=731166
There's no need for anyone to poke at the libvirt bridge that's not an
admin. It should probably be locked down more as well, but that would be
NetworkManager's business.
Since we now re-sort the model according to the search terms, we can't
use tree model paths as results since they're not stable. Instead
we'll use panel IDs and keep a map of IDs to GtkTreeIters to retrieve
the result metas from the model.
Caching GtkTreeIters in this way is OK because the model is a
GtkListStore which guarantees that while a row exists, the iter is
persistent.
https://bugzilla.gnome.org/show_bug.cgi?id=729027
It's a likely word that might be used when looking for this panel and,
with the new search results ordering, the power panel is coming up
ahead due to it preferring keyword matches when they exist.
https://bugzilla.gnome.org/show_bug.cgi?id=729027
Before, we were freeing the DateTime instances when closing the dialog
through the close button and always creating new ones on _show().
Now, with the header bar, we go through the delete-event handler which
just hides the dialog, meaning that we leak the dates instanced on
_show().
Clearing the dates on _show() is more robust anyway. Note that we'd
leak even before switching to a header bar when going through the
delete-event handler.
https://bugzilla.gnome.org/show_bug.cgi?id=729181
Instead of just sorting by the name the sort order will now be:
1. Panels whose name match a search term
2. Panels whose keywords match the most search terms
3. Panels whose description match the most search terms
4. The remaining panels by name
https://bugzilla.gnome.org/show_bug.cgi?id=729027
