[gdal-dev] MSYS buildslave

[Even Rouault]

Ari,

fopr the crash in MEMRasterBand::IReadBlock(), I would suspect that 
CPLScanPointer doesn't do its job rightly when compiled in MSYS environment.

The function looks currently like :

/* -------------------------------------------------------------------- */
/*      On MSVC we have to scanf pointer values without the 0x          */
/*      prefix.                                                         */
/* -------------------------------------------------------------------- */
    if( EQUALN(szTemp,"0x",2) )
    {
        pResult = NULL;

#if defined(WIN32) && defined(_MSC_VER)
        sscanf( szTemp+2, "%p", &pResult );
#else
        sscanf( szTemp, "%p", &pResult );
#endif
    }

I'd guess we probably don't need the defined(_MSC_VER) test


Le Saturday 01 November 2008 17:29:17 Ari Jolma, vous avez écrit :
> The wrappers seemed to build with Python 2.4 without other changes than
> uncommenting libraries = gdal in setup.cfg. setup.py seems to set gdal_i
> by default in windows.
>
> The autotest dumps core in usgsdem.py test 4 (the culprit is memcpy in
> MEMRasterBand::IReadBlock), but I'm not sure if I have the most recent
> gdal right now.
>
> These are not yet in the buildslave, which currently is failing because
> I have something running there which keeps the swig/python directory
> undeleteable - damn windows!
>
> Ari
>
> Christopher Barker kirjoitti:
> > Andrey Kiselev wrote:
> >> Otherwise of that bug Python stuff should be perfectly compilable and
> >> usable with the official binary Python distribution.
> >
> > At least with python2.5 I'm not sure about 2.6, but earlier pythons
> > required some patching to work with MingGW binaries.
> >
> > -Chris