[osgeo4w-dev] Problems in using Perl wrappers with osgeo4w distro

[Tamas Szekeres]

BTW: Is it possible to include multiple versions of the binary
packages in osgeo4w side by side? I confirm that the compilations
created by different compilers don't work very well together and we
may encounter strange effects because fo this.
What I would really like to see is that from a particular SWIG package
we could refer to the desired versions of the dependencies.
Distributing multiple versions of the same binary package regarding to
the compiler versions in not uncommon practice on Windows.


Best regards,

Tamas



2008/3/16, Frank Warmerdam <warmerdam at pobox.com>:
> Ari Jolma wrote:
>  > Hello,
>  >
>  > I'm able to build Perl wrappers from GDAL 1.5 against the latest osgeo4w
>  > binary release. I use MinGW (g++ is 3.4.5 with which my perl 5.10 is
>  > also built). The link library is gdal_i.lib - it just needs to be
>  > renamed to gdal.dll.a for perl to find and use it.
>  >
>  > The module loads ok, and calls some functions ok from gdal15.ddl but
>  > fails when calling CPLError. Debugging the ogr_wrap.cpp I see that
>  > OGR_G_CreateFromWkb is called ok (but it returns deliberately an error),
>  > OGRErrMessages(err) gives me correct error message, but then calling
>  > CPLError(CE_Failure, err, "%s", OGRErrMessages(err)); never returns and
>  > Windows pops up the "has encountered a problem" dialog. gdb says that
>  > there is a Segmentation fault and the stack looks corrupt.
>  >
>  > I wonder what could be the problem and how to even study what's the real
>  > problem.
>  >
>  > With what compiler is gdal15.dll built?
>
>
> Ari,
>
>  gdal15.dll is built with VC++ 7.1 (aka Visual Studio 2003).  I have skimmed
>  this in the perl bindings:
>
>    OGRGeometryShadow* CreateGeometryFromWkb( int len, char *bin_string,
>                                              OSRSpatialReferenceShadow
>  *reference=NULL ) {
>      void *geom;
>      OGRErr err = OGR_G_CreateFromWkb( (unsigned char *) bin_string,
>                                        reference,
>                                        &geom,
>                                        len );
>      if (err != 0 ) {
>
>         CPLError(CE_Failure, err, "%s", OGRErrMessages(err));
>
>         return NULL;
>      }
>      return (OGRGeometryShadow*) geom;
>    }
>
>  I don't see any obvious likely issues, such as allocations and frees
>  taking place out of different heaps.  Is it possible that var-args lists
>  don't work compatibly between gcc and vc++?  If this turns out to be the
>  case we may need to provide a non-varargs based error function.
>
>  I would note that OGRErr values are not really expected as the second
>  argument to CPLError().  You might almost be safer passing CPLE_AppDefined
>  in this situation, though I can't imagine that having any significance
>  in this case.
>
>  Lastly, you might want to confirm that OGRErrMessages(err) is non-NULL.
>
>  Best regards,
>
> --
>  ---------------------------------------+--------------------------------------
>  I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
>  light and sound - activate the windows | http://pobox.com/~warmerdam
>  and watch the world go round - Rush    | President OSGeo, http://osgeo.org
>
>
>  _______________________________________________
>  osgeo4w-dev mailing list
>  osgeo4w-dev at lists.osgeo.org
>  http://lists.osgeo.org/mailman/listinfo/osgeo4w-dev
>