[Gdal-dev] About the VRTWarpedRasterBand subclass

[Frank Warmerdam]

Mario Beauchamp wrote:
> I was not aware of those issues. This is actually in Msys, a mingw bash for win32. Anyway, same
> result in a DOS window.

Mario,

The problem is that internally I actually compare the transformer function
pointer to determine how to serialize the transformer.  But the pointer to
a function will vary depending on whether you are "in the DLL" where it is
defined, or outside the DLL.  That relates to the way that DLLs function call
indirection is handled.  Apparently you actually link in a stub function into
your calling .exe (or .dll) and then this stub function looks up the real
function and calls it.  This makes my whole approach to serializing transformer
fall apart unless you happen to statically link one big giant executable (or
form your GDALWarpOptions within the GDAL DLL itself I suppose).

As far as I know this would affect MSYS built software just the same as it
would affect Visual C++ built apps.

> I am not sure it's a bug now... I guess the question should be: is the VRTWarpedRasterBand fully
> operational? If so, then my problem is the VRT serialization on windows issue.

Please file the bug in bugzilla, and add my explanation up above into the report
as well so I will remember the details when I get to fixing it.  In the
meantime, I *think* the VRT warper serialization stuff is essentially
broken on Windows.  I will likely need a fundamental rewrite of how transformers
are encapsulated to make this work properly.

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    | Geospatial Programmer for Rent