Re: segfaults with Python 2.3.x

Well, that's odd, because 2.0.5 was released because 2.0.4 was
incompatible with pymalloc, and had segfaults with 2.0.4 since the
earliest 2.3 betas.  Do you have python compiled with pymalloc turned off?

As it happens, I've been using 2.0.5 mostly, but chose to patch 2.1.0
for some reason.  The same problem was taking place regardless.	

Incidentally, this seems to be a "watched pot never boils" issue.  I
got up this morning to see that several segfaults had occurred during
the night.  A core file shows malloc dying in time_strftime, in the
time module.  Odd.

On Thu, Feb 12, 2004 at 06:13:38AM -0800, Drew Csillag wrote:
> I've been using egenix mx-base 2.0.4 and have not been having any
> core dumping problems at all.  So if you're having problems with
> 2.1.0 (which it appears you are using), 2.0.4 may be the way to go.
> 
> drew@loki:~/build>~/local/stow/Python-2.3/bin/python
> Python 2.3 (#1, Aug 12 2003, 13:04:46) 
> [GCC 2.96 20000731 (Red Hat Linux 7.1 2.96-81)] on linux2
> Type "help", "copyright", "credits" or "license" for more
> information.
> >>> import mx
> >>> dir(mx)
> ['__builtins__', '__copyright__', '__doc__', '__file__', '__name__',
> '__path__', '__version__']
> >>> mx.__version__
> '2.0'
> >>> 
> 
> Cheers,
> Drew
> 
> --- Jacob Smullyan <[email protected]> wrote:
> > On Wed, Feb 11, 2004 at 05:05:07PM -0500, Jacob Smullyan wrote:
> > > DUH.  Patch now attached.
> > > 
> > > I may have jumped the gun in any case.  I'm not absolutely
> > confident
> > > that this is an adequate solution; I just found a segfault on a
> > server
> > > I thought I had patched.  More later.
> > 
> > I am now pretty sure that this patch is not a complete solution,
> > but
> > my tests so far seem to indicate that skunkweb is segfaulting much
> > less than before -- over the sample tested today, an order of
> > magnitude fewer segfaults (but not none) on a complex site with
> > several other C extensions also loaded that could also be
> > responsible
> > for problems.  That being said, it is quite possible that this is a
> > statistical fluke, or that other factors affected the results.
> > 
> > I'm attaching the patch I'm using at present, which turns off
> > another
> > bulk memory management mechanism in mxDateTime.c.  Let me know if
> > it
> > helps, or doesn't ....
> > 
> > js
> > 
> > > *** mxDateTime.c	Wed Feb 11 22:17:15 2004
> > ---
> >
> /home/smulloni/src/egenix-mx-base-2.1.0-real/mx/DateTime/mxDateTime/mxDateTime.c
> > Sat Aug  9 23:12:34 2003
> > ***************
> > *** 77,86 ****
> >   
> >   /* Define these to have the module use free lists (saves malloc
> > calls) */
> >   #define MXDATETIME_FREELIST
> > - /* modified by js Wed Feb 11 11:28:06 2004 */
> > - #undef MXDATETIME_FREELIST
> >   #define MXDATETIMEDELTA_FREELIST
> > - #undef MXDATETIMEDELTA_FREELIST
> >   
> >   /* Define this to enable the copy-protocol (__copy__,
> > __deepcopy__) */
> >   #define COPY_PROTOCOL
> > --- 77,83 ----
> > 
> 
> > ATTACHMENT part 2 application/pgp-signature 
> 
> 
> 
> __________________________________
> Do you Yahoo!?
> Yahoo! Finance: Get your refund fast by filing online.
> http://taxes.yahoo.com/filing.html
>

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)

iD8DBQFAK5dHuqamFyFXXLIRAomWAJ9gQn5ZipfXN7zDALXS7ZY7+TZ/eACfcnIf
5N4kVwxgotrZtuLZTTDeSOc=
=QzcH
-----END PGP SIGNATURE-----