Re: [Pypgsql-users] segfaults

[Gerhard H. <gh...@gh...>]

Craig Main wrote:
> On Wed, 2003-04-23 at 14:47, Gerhard Häring wrote:
> 
>>Ok, so it doesn't actually die in the module initialization phase of the 
>>C module. Could you please add print statements into 
>>pyPgSQL/libpq/__init__.py so to show where it dies there?
>>
>>(This sounds like an obscure bug to me, perhaps caused by too agressive 
>>compiler settings that trigger a gcc bug.)
> 
> Good news (maybe) for the pypgsql team, I seem to have tracked the
> problem down to the "from mx import DateTime" line in the PgSQL.py file.
> Running the that line from the python shell segfaults. So it appears
> that the root cause of the segfaulting is not the pypgsql code. The line
> "from pyPgSQL import libpq" does not crash...
> 
> Any ideas? 

If you've compiled your Python with --with-pymalloc, mxDateTime will 
crash. IIRC this has been fixed in the latest mxExtensions beta.

I'd recommend that you recompile Python and mxDateTime with safe 
settings (no -O3, --with-pymalloc, -march=athlonxp or any such fancy 
things). If it's not a mxDateTime bug, it's likely that you've hit code 
generator or optimizer bug in your compiler. This happens more 
frequently than you'd probably think.

>I dont seem to find a mx-dev list anywhere?

http://www.egenix.com/mailman/listinfo/egenix-users

-- Gerhard