[AG-TECH] AG2.1.2 Certificate Manager with python2.3

Ivan R. Judson judson at mcs.anl.gov
Wed Jan 21 17:44:17 CST 2004


I did get Keith to commit to putting these interfaces in pyGLobus today, so
if we can get them summarized for Matt he can blow through them next week.

--Ivan 

> -----Original Message-----
> From: Robert Olson [mailto:olson at mcs.anl.gov] 
> Sent: Wednesday, January 21, 2004 5:38 PM
> To: Christoph Willing
> Cc: Ivan R. Judson; 'AG-TECH'
> Subject: RE: [AG-TECH] AG2.1.2 Certificate Manager with python2.3
> 
> I may reexamine how the extension stuff is done in there; 
> I've been poking around at that code today, as a matter of 
> fact, for a different reason.
> 
> I'll keep you posted, and will try to reproduce your crash.
> 
> --bob
> 
> On Thu, 22 Jan 2004, Christoph Willing wrote:
> 
> > On Wed, 2004-01-21 at 16:46, Robert Olson wrote:
> > > are you using openssl 0.9.7 by any chance? this sounds 
> like an open 
> > > issue we have with pyOpenSSL and that version of the openssl libs.
> > > 
> > > --bob
> > > 
> > 
> > 
> > On a system with python2.2, we have openssl-0.9.7b and all is OK.
> > 
> > On a system with python2.3, we have openssl-0.9.7c and 
> Cert. Manager 
> > does a seg fault trying to do:
> > 	xext = crypto.X509Extension(name, critical, value) at 
> line 604 of 
> > CertificateRepository.py.
> > 
> > If I additionally install python2.2 in this second system, then the 
> > Certificate Manager runs OK again. This makes me suspect the python 
> > version.
> > 
> > 
> > 
> > While googling for possible clues, I came across 
> > http://mail.python.org/pipermail/python-list/2003-July/174843.html
> > - unhelpful resolution, but that email questions using pyOpenSSL 
> > (where I think the crypto class lives), when the python2.3 socket 
> > module already has SSL support built in.
> > 
> > 
> > Also, I see in pyOpenSSL_AG-0.5.1/src/utils.h that pymemcompat.h is 
> > included, which is supposed to massage some different python memory 
> > APIs. Could these differing memory APIs still be problem 
> despite the 
> > attempt at compatibility? - the crypto.X509Extension90 call 
> produces a 
> > seg fault...
> > 
> > 
> > chris
> > 
> > 
> > 
> > 
> > 
> 
> 




More information about the ag-dev mailing list