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.

The idea is good, but seems that error handling should be inlined into initiate_send.

Also those 3 special exceptions should be defined once in the module instead of repeating them.

This issue is 6 years old. I don't understand the bug and the patch has no unit test. I suggest to close the issue as out of date. The asynchat module is now deprecated in favor of the new shiny asyncio module.

#6550 and #11267 have been closed as out of date as asynchat and asyncore are effectively deprecated in favour of asyncio so can't this go the same way?

asynchat has been deprecated for a long time, it's unlikely we will do anything with it now.