[osgeo4w-dev] refactoring grass code for MSVC: is an estimate possible?

[G. Allegri]

Dear all developers,
in these days me, and other users, are facing serious problems with
the actual status of grass (and qgis-grass plugin) over the Windows
Vista system. I've spent two days the give an insight in the structure
of the build environment for Windows, MinGW and MSVC, and with the hep
of Jef I've tried to solve some of the first problems (crashes) I was
facing, mostly due to memory managements issues.
I know that there's been a loft of talks and thread on the argument,
and trying to sum up the conclusions it appears that the best
improvement would be a profound code refactoring for grass to enable
its building with MSVC, since MSVC compiled code performs better then
MinGW and, as far as I have seen, MinGW/MSVC coupling gives unstable
results.

The question is: how much effort (time and money) would require such
an operation?
If it results unfeasable, what should/could be done to make a more
stable structure (also meant as continuity in time) to the grass and
grass-plugin builds for Windows (inside or outside OSGeo4W)?

If we could make an economic estimate of the job to be done, we could
try to organize a found raising dedicated to it. I would be the first
to contribute.

Thanks to all of you, for everything that has been done till now and
for what will be done from now on.
Giovanni