Previously the function returned the distance to the nearest item. Now
it only returns an item that is hit. This slightly changes semantics
(button events are no longer dispatched to the nearest item, but only to
the item actually clicked on), but makes the code way simpler and
actually does what one would expect.
In GTK+ 2.21.8, the keysym names were renamed from GDK_* to GDK_KEY_*.
I've added backward-compatibility macors to gtk-compat.h, which can be
dumped as soon as we require GTK+ >= 2.22.0.
Work around the issue of GnomeCanvasItem amending its own flags to
GtkObject::flags (which is sealed) by giving it its own flags field.
This breaks libgnomecanvas ABI and API, but I see no other way.
This commit didn't work the first time because gnome-pilot libraries
were still pulling in the system-wide libgnomecanvas, and that was
interfereing with our bundled version which has a different ABI.
But gnome-pilot integration was dropped in the previous commit, so
everything is now using the bundled libgnomecanvas.
This reverts commit fd8b55edaa.
Something in this commit seriously hosed ETable, making Evolution pretty
much unusable. Reverting this until I can track down the problem.
Work around the issue of GnomeCanvasItem amending its own flags to
GtkObject::flags (which is sealed) by giving it its own flags field.
This breaks libgnomecanvas ABI and API, but I see no other way.
To avoid another case like bug #587014, add GSignalAccumulator functions
to all ETable and ETree signals that return a flag to indicate the signal
has been handled.
See commit e9dc381d3a for an example of
the kind of problems not having a GSignalAccumulator can cause.
Signals changed:
ETree::click
ETree::right-click
ETree::white-space-event
ETable::click
ETable::key-press
ETable::right-click
ETable::start-drag
ETable::white-space-event
ETableItem::click
ETableItem::right-click
ETableItem::start-drag
ETableGroup::click
ETableGroup::key-press
ETableGroup::right-click
ETableGroup::start-drag
Neither ETree::key-press nor ETableItem::key-press signal definitions
specify a GSignalAccumulator that terminates the signal emission when
a handler returns TRUE, and Evolution 2.29 connects multiple handlers
to ETree::key-press.
When the space key is pressed, the first handler implements the magic
space bar behavior and returns TRUE, which should terminate signal
emission but doesn't because there's no accumulator function on the
signal. So the second handler runs and checks for other keys besides
space. It returns FALSE since it didn't handle the key press.
End result: emission site gets back FALSE (from the second handler) as
the return value, so it thinks the key press was not handled at all and
invokes the fallback handler -- e_selection_model_toggle_single_row() --
which -unselects- the newly selected row, making it appear the next
unread message was never selected.
Several other ETable-related signals that return "handled" flags also
lack GSignalAccumulator functions. I've made a note to myself to fix
those as well.
- Collect all shell modules into a new top-level 'modules' directory:
$(top_srcdir)/modules/addressbook
$(top_srcdir)/modules/calendar
$(top_srcdir)/modules/mail
Nothing is allowed to link to these, not plugins nor other modules.
THIS SOLVES BUG #571275 AND OPENS THE DOOR TO PORTING TO MAC OS X.
- Mimic the libevolution-mail-shared library from master (except drop
the "shared" suffix) and have libevolution-mail-importers and all
mail-related plugins link to it.
- Discard the a11y subdirectories and have the files live alongside
their counterpart widgets.
