gimp_drawable_get_mode_node() was calling gimp_drawable_get_node()
instead of gimp_item_get_node() if the mode_node didn't already exist.
Fortunately that was never the case i nthe current code. Added
warnings in the GimpItem::get_node() impls that would go off in such a
situation.
Make gimp_ui_get_ui_manager() in test-ui.c handle when there is no
empty display so we can put the new-image test before the
single-window mode tests. We want the single-window mode tests lasts
because they are so flickery.
and change the dialog from a editor of the display's devices to an
editor of all devices of all displays known to GIMP, regardless
whether currently plugged or just remembered in devicerc. will need to
distinguish them in the next step...
because the device info container wants to be displayed in a container
view soon. Also we are about to get lists of tool presets and stuff,
which are also contexts.
Add functions that are similar to the "device-added" and
"device-removed" of GdkDeviceManager in the XI2 branch. Add "device"
and "display" properties to GimpDeviceInfp and distunguish between
creating device info objects from scratch for never-seen devices, and
setting a GdkDevice on them when a previously known device gets
plugged in.
Added two new widgets, GimpDeviceInfoEditor, which is an editor/view
widget for GimpDeviceInfo, and GimpDeviceEditor, which is an editor
widget for all devices. Both are pretty much ugly right now and look a
lot like the old GtkInputDialog, but are at least internally cleaned
up and easily changable code and ui wise. Consider this a completely
intermediate state.
Also cleaned up GimpDeviceInfo so it's possible to have a proper
view on it, and did the needed changes to the preferences dialog
to use the new stuff.
Move calling gimp_plug_in_shm_free() from gimp_plug_in_manager_finalize()
to gimp_plug_in_manager_exit() so it's also called in stable GIMP
releases which simply call exit() at some point instead of taking the
time to shut down everything completely.
For GIMP_UNSTABLE, keep around a list of allocated tile managers and
have a function tile_manager_exit() which complains about them and
unrefs them. This is infinitely more helpful than the tile cache and
swap complaining about not being empty, because there is absolutely
nothing wrong with swap and cache when we simply leaked tile managers.
Add gimp_image_new_from_drawable(), from_component() and from_pixbuf()
and remove that duplicated code from gimptoolbox-dnd.c and
gimpdisplayshell-dnd.c