[IPython-user] Towards IPython 1.0, the famous big cleanup

Fernando Perez Fernando.Perez at colorado.edu
Thu Apr 14 23:34:08 CDT 2005

John Hunter wrote:

> While I respect the "there's only one Fernando" argument, I'm lobbying
> for a branch.  It's not too hard to keep a bugfix release tree and if
> someone sends a mission critical bug against that tree to incorporate
> it and release it.  I think you can free up most of your ipython time
> by doing a feature freeze, while still supporting critical bug-fixes
> against the old tree.
> Maybe some kind ipython user should pony up and offer to maintain a
> bugfix tree of the 0.6.13 tree?  Basically, all you have to do is
> apply critical patches and pester Fernando to upload the bugfix
> releases every 4 weeks or so until he emerges from ipython 3000.  I
> would volunteer to do it but, like Fernando, am over-committed on this
> crazy open source venture myself.  Volunteers?

Not to worry... Jack Moffit (debian maintainer) already mentioned this, so 
I'll just work on HEAD, but right before starting, I'll cut a MAINT branch. 
This will leave a playground for critical bugfixes to be applied with minimal 
pain.  I guess I'll have to actually read the damn CVS info manual to figure 
out branching, but that's my problem, not anyone else's :)

> I got a segfault on tab completion the other day in CVS ipython but
> could not replicate it.  Hit the tab key and boom, everything was
> gone.  I was using wx I think.  Tried reloading the libs and
> re-executing the same commands but couldn't replicate it so I didn't
> report it.  My point is that the gthread and wthread code is still
> potentially flaky and serious, mission critical bugs could still be
> lurking.

Yes, that threading code is definitely a minefield.  Also note that readline 
completion does eval() calls often (poke around rlcompleter in the stdlib or 
FlexCompleter in ipython); with libs like wx/gtk, this can be a recipe for 
easy segfaults.  Since ipython itself is pure python, short of bugs in the 
python core or stdlib it should never be able of segfault by itself.

If you can ever reproduce this, send it my way.  However, it is not impossible 
that the bug is in readline; an ipython user recently found such a bug:


It turned out to be a readline problem (this _might_ have been what you saw) 
and he eventually followed it up all the way to the Python SF tracker, where 
it was later fixed by M. Hudson:


Thanks for the comments!



More information about the IPython-user mailing list