Re: [python] Crash when using generators with Python 2.4

[Phil Muldoon <pmuldoon at redhat dot com>]

Paul Pluzhnikov wrote:

Hi Paul,

> On Tue, Feb 24, 2009 at 8:19 AM, Phil Muldoon <pmuldoon@redhat.com> wrote:
>
>   
> > Before I start, I'm curious what you were observing? Do you have any data
> > that you can publish on this?
> >     
>
> The valgrind output was attached to my Jan 20th message in this
> very thread:
> http://sources.redhat.com/ml/archer/2009-q1/msg00095.html
> http://sources.redhat.com/ml/archer/2009-q1/txt00001.txt
>   


For some reason in my mailbox, Thunderbird did not have that 
attachment.  Weird.  Anyway thanks for the web-link.

When I run some pretty-printers that pretty-print: deques, bitsets, 
vectors  and classes I see about 7 Kbytes of loss.

> This output will likely not be very useful now, since quite
> a bit of code has changed. One thing I did to reduce VG noise
> was build python (to be linked into GDB) with --without-pymalloc.
>   


Thanks for the tip on without-pymalloc.  I did try building Python with: 
--with-pydebug, to look at the reference counting debug feature.  It 
seems it is being swallowed somewhere in the output to GDB.  Have not 
had time too look at it in more detail yet.

> Let me know if you want me to rerun my VG experiments with
> current sources

If you have some time to do that, sure,  that would be a big help.

Thanks

Phil