Suppose you have tow sockets handled by two asynchat.async_chat class
instances and from class1 you want to send some data through class2:

class1(asynchat.async_chat):
    
   ...
   
   def mymethod(self):
       class2_instance.push_with_producer(producer)


Since push_with_producer automatically calls initiate_send() if an error
occurs while class2 reads the first chunk of data from the producer,
class1's handle_error gets called.
If the error occurs when reading the next chunks of data class2's
handle_error will be called instead.
This is an inconsistency that should be fixed so that always class2's
handle_error gets called.

The patch in attachment does that.