[PEAK] access violation with py24/msvc compiler

alexander smishlajev alex at ank-sia.com
Fri Jan 28 08:32:28 EST 2005


Phillip J. Eby wrote, at 27.01.2005 23:02:
> 
>>>> thanks, i did it now.  it is kjGraph.  the following command also 
>>>> raises access violation:
>>>>
>>>>    python -c "import kjbuckets; g=kjbuckets.kjGraph(); print g"
>>>>
>>>> when importing peak.model.features, failing is line 470 in
>>>> method_exporter.py: 'vt += d' in MethodExporter.__init__()
>>>>
>> as i said earlier, kjGraph object raises access violation when i 
>> attempt to print it.  so in this case AV was caused by pdb trying to 
>> print result of the statement evaluation, and not by addition.

this statement was wrong.  addition raises AV too, but that depends on 
too many circumstances.  i've seen that command

   python -c "from peak.model import features"

stops segfaulting when i add 'import pdb' to the body of class structField.

> Ah.  Okay.  Try removing the printfunc from the type structure then, and 
> see how it works.  (Presumably it will fall back to printing the str or 
> repr).  If it doesn't crash, then, we'll know that it's a file pointer 
> issue.

it fails in function newWrapper, in the call to PyMem_NEW at line 1724.

by the way, it seems to me that PyMem_Del(wp) is missing before return 
NULL at line 1737 in that function.

in fact, i would like to stop this investigation.  mingw build is ok for 
us.  msvc would be more handy, but that is not worth the time spent.

thanks for your help.

best wishes,
alex.




More information about the PEAK mailing list