Re: [Matplotlib-users] Re: [SciPy-user] ANN matplotlib-0.30 - matlab style python plotting

[John H. <jdh...@ac...>]

>>>>> "Barry" == Barry Drake <bl...@ad...> writes:

    Barry> I'm observing some difference in behavior between the two
    Barry> versions.  Running some example code in 0.29, after closing
    Barry> the plot windows, the python.exe process quits normally.
    Barry> The same example using 0.30 hangs the process and I have to
    Barry> kill it by hand.  This occurs both at the command line and
    Barry> from within my editor (normally I execute from the editor).

    Barry> Any ideas?  Do I need to change some set up from when I was
    Barry> using 0.29?

Yes.  I made several changes to this part of matplotlib to iron out
some interactive bugs, and a new one crept in; I didn't notice it in
my tests.  The quick fix is to add the following code to
matplotlib/backends/backend_gtk.py at line 773

        if GcfGTK.get_num_figwins()==0 and ShowOn().is_mainloop_on():
            gtk.mainquit()

This is in the GcfGTK.destroy  function, so the entire function should
read

    def destroy(num):
        if not GcfBase.has_fignum(num): return 
        figwin = GcfGTK.figs[num]
        figwin.figure.drawable= None
        GcfBase.destroy(num)
        if GcfGTK.get_num_figwins()==0 and ShowOn().is_mainloop_on():
            gtk.mainquit()
    destroy = staticmethod(destroy)

This should restore the old functionality -- let me know if you see
anymore strangeness.  I'll check the change into CVS.

[BTW, I posted the 0.30 announce to this mailing list this afternoon,
but it didn't seem to make it.  Hmmm....  Have I been moderated out of
existence :-) ]

Thanks for the bug report, 
John Hunter