Tkinter -- interpreter crashes in tcl in presence of threads?

Bernard Sufrin sufrin at comlab.ox.ac.uk
Thu Aug 16 12:15:42 CEST 2001


Thanks for your contribution. I agree that it's not a python coding
problem. 

Sadly, the newest stable Tk installations warn about unreliability of Tk
in a multithreaded environment (particularly when linked with glibc). 

My solution was, in the end, more radical: I replaced the GUI objects in
the event simulator application with proxies that communicate with an
outboard singlethreaded ``GUI server'' process via a socket (unix or
inet domain depending on operating system). This degrades
interface-to-application bandwidth somewhat, but not so much that it
matters.

The positive thing about the exercise (it lasted half a day) was  that
it forced me to make explicit the spec of the interface between
application and GUI. A consequential bonus was that it was then easy to
develop a multi-user, multi-machine GUI for the application....

As for the proof assistant interface, it was relatively easy (2 hours)
to convert it to a form where it's driven by events (including requests
for input) sent from the proof engine.

Bernard Sufrin.



More information about the Python-list mailing list