Hi Miguel, Sorry for the long delay. After more thinking, I realized that your original approach (splitting the parser for Glade 1, Glade 2, and GtkBuilder) would be better for the future maintenance, like this patch:
https://github.com/644rosen/gettext_gtkbuilder_support/commit/19f6a2c6 although I don't like some part, particularly kind of micro-refactoring like this: static void init_glade2_tags () { /* When adding new keywords here, also update the documentation in xgettext.texi! */ static const char *tags_array[] = { "property", "atkproperty", "atkaction", NULL }; init_tags (tags_array); } In do_glade2_start, you still need to manually check the element name for checking the description="..." attribute of the <atkaction>. This looks like a refactoring failure :-) Anyway, I've pushed a revised version and the test case. Thanks very much for the patches. Regards, -- Daiki Ueno
