I'd like to read/write from/to a registered dictionary, but I have some trouble understanding how dictionary accessing is managed in a multi-threaded scenario. So, basically there are two separate questions here:
(1) Should dict-related function calls always be deferred to the main thread (and can developers rely safely on this assumption, ie. are all dict-related standard objects implemented with this taken into consideration)?
The following question, I guess, makes only sense if the answer to (1) is 'no':
(2) When I make a call to dictobj_findregistered_retain(), is the dict locked against further modifications as long as I don't make a call to dictobj_release(), or should I always call dictobj_findregistered_clone() in the case when I need to read a lot from the dict?
More generally: if the answer to (1) is 'no', then how are we supposed to properly read and write dictionaries?