[Python-modules-team] Bug#448530: more info

[Jan Medlock]

On Friday 14 December 2007 02:58:58 pm Ondrej Certik wrote:
> Hi Jan,
>
> could you please send us more information about this bug? We would like
> to reproduce it and fix it. See my previous email about that, or browse
> it online here:
>
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=448530
>
> Thanks a lot,
> Ondrej

Hi Ondrej,
For some reason, I didn't get the previous message to the bug number...

It looks like you have a garbled version of the pickle file.  I've attached it 
here again, base64-encoded this time.

$ md5sum phi.p
73b409eed67ce1eef0d0c1f0c974cbcd  phi.p

Also, for some reason, after the latest update of Numpy a week or so ago, 
LD_LIBRARY_PATH no longer changes which ATLAS library gets loaded in this 
case.  I now must use LD_PRELOAD.

$ LD_PRELOAD=/usr/lib/atlas/sse2/libblas.so.3 python test.py
About to multiply
Segmentation fault

$ LD_PRELOAD=/usr/lib/atlas/sse/libblas.so.3 python test.py
About to multiply
Multiplied

I'm happy to help track this down.  Sorry my example is so complicated---it's 
as simple as I could get it that would reproduce the bug.

Thanks for looking into this.
Jan
-------------- next part --------------
A non-text attachment was scrubbed...
Name: phi.p
Type: application/octet-stream
Size: 6455 bytes
Desc: not available
Url : http://lists.alioth.debian.org/pipermail/python-modules-team/attachments/20071214/cb0006c9/attachment.obj