Created attachment 27780 [details]
List o' leaks

A few of these are the fault of orbit.

Summarizing the leaks:

Fixed:
persistence_save_list never freed the string of a GString, fixed.
edit_delete_callback alloced a list that didn't get copied, fixed.
dia_object_defaults_save used NULL for key free function, fixed.
recent_file_menuitem_create does string manip, uncertainty about allocation,
fixed but needs testing.
parent_list_affected_hierarchy allocs a hash table, fixed.

Not yet fixed:
dia_toggle_button_new_with_images allocs two images that are not freed.  Does
unref, but maybe should sink rather than ref?
object_load_props gets a prop list back, when freed the font_prop isn't freed?
diagram_tree_add copies a glist into g_list_concat, the result is freed, though.
dia_font_scaled_string_width calls pango, and something arabic is never freed. 
I believe we have another outstanding bug about this.
gtk_init stuff is not our fault.

dia_toggle_button_new_with_images was probably that buttons not clicked didn't
have one of the images sunk, now sinking.

Sure there are leaks left, but these appear to be fixed now.
dia_toggle_button_new_with_images is gone

2005-06-18  Hans Breuer  <hans@breuer.org>

	[plugged remaining leaks from bug #142669]
	* lib/prop_attr.c(fontprop_free) : not only unref the font
	but free the property itself, too. 
	* app/diagram_tree.c(create_diagram_children): remember the
	original list start to not only free the last element 

	* lib/prop_sdarray.c:64: warning: `arrayprop_free' defined 
	but not used. Here it meant a potentially huge leak.