this patch does two things to asyncore.py, none of them should break any existing code which uses the module:

1) remove the (broken) poll2 function, and rather make a reference from poll2 to poll3 in the module namespace (i.e. if someone actually uses the poll2 function, poll3 will be called instead).

2) add a function loop_once which performs the asyncore loop once instead of until the socket map is empty.  asyncore.loop_once has the same function signature as asyncore.loop.

Logged In: YES 
user_id=21627

Can you please explain in what way poll2 is broken?

Also, what happens if use_poll is set, but there is no poll 
available on the system?

Logged In: YES 
user_id=5685

asyncore.poll2 is broken because it does not mask the EINTR exception when invoking the poll.poll function.  this causes asyncore.poll2 to not have the same semantics as asyncore.poll and asyncore.poll3.

i don't see the point in preserving support for the poll2 function, as the poll function is now automatically a part of the select module if the system has poll support.   or am i missing something obvious here? :)

if use_poll is set and there is no poll on the system, asyncore.loop and asyncore.loop_once will fail the same way they did before.  

Logged In: YES 
user_id=11375

poll2() is written for a "poll" extension module that came with the Medusa 
distribution, so it'll never be used with a Python installation and poll2() can be 
safely removed.  

I'm less enthusiastic about loop_once().  Existing code that needs to loop once 
has been simply calling asyncore.poll(), which works fine (though perhaps 
suboptimally -- but 99% of asyncore applications won't care).  

Logged In: YES 
user_id=11375

Partially checked in; the existing poll2() implementation has been removed in 
rev.1.43 in CVS.

Logged In: YES 
user_id=11375

loop() now takes an optional 'count' parameter, so this patch can now be 
closed.  Thanks for submitting it!