[IPython-user] 96% cpu usage with new scipy

Fernando Perez Fernando.Perez at colorado.edu
Mon Dec 26 02:38:26 CST 2005

Ryan Krauss wrote:
> I don't have QtAgg.  TkAgg and WXAgg seem to be fine.  It seems like
> only GTKAgg has this problem.  I went with that one because I am
> running gnome.

I'm getting the feeling that the GTK* backends really have a problem: there 
currently is another user on the mpl list reporting strange errors there only 
with GTKAgg, and I've seen it segfault repeatedly on my box.  It might be that 
all these are ultimately the same problem, but unfortunately it seems it's 
something subtle, low-level, threads-related and with other dependencies (as 
it doesn't appear on all OSes or even on the same box depending on whether the 
session is local or remote X11).

I'm afraid I really won't be able to sink much time into this one: I don't 
know the mpl/GTK codebase at all, and I'd have to spend a lot of effort here 
before I could really make any progress.  Given that I'm trying to at least 
pick up momentum again with ipython itself, I'll have to punt on this one and 
leave it to John or another of the mpl devs.  Given that ipython only makes 
high-level, pure python calls to initialize the GTK threading engine, I really 
suspect that this is an mpl bug (or even pyGTK one), however one that is only 
exposed when ipython sets up the threading machinery.

Sorry not to be of more assistance on this one.



More information about the IPython-user mailing list