[GRASS-dev] WinGRASS Wiki Hacking

[Glynn Clements]

Frank Warmerdam wrote:

> > Sooner or later, someone will need to bite the bullet and produce a
> > dependency-based installer, probably based upon Cygwin's.
> > 
> > The only question is how much time gets wasted on half-baked solutions
> > between now and then.
> 
> Glynn,
> 
> I think that's what I'm doing already with OSGeo4W!  It includes
> a dependency based installer derived from the Cygwin one (but with
> no dependency on cygwin runtime dlls).  It is intended to provide
> an integrated "production ready" windows FOSS4G solution for all
> OSGeo projects, and other projects of interest.
> 
> There is already a working version available, though it is rough
> around the edges, and the set of packages is still incomplete.
> 
> It includes a somewhat half-baked WinGRASS build I prepared myself
> several months ago, based on the wingrass build notes at the time.
> Amoung other things, it proves that a MingW built GRASS can use
> a VC7.1 built GDAL.
> 
> I've been mildly disappointed in the past that Marco wasn't
> interested in preparing a WinGRASS package for OSGeo4W, but
> he explained that it seemed like more than he wanted to take
> responsibility for at the time.  I'm hopeful that someone else
> from the GRASS community will take on this responsibility.
> 
> More info available at:
> 
>   http://trac.osgeo.org/osgeo4w

I downloaded the installer and, yes, that's exactly what I was
suggesting that GRASS should be doing. It shouldn't be necessary to
re-download dozens of third-party libraries with every GRASS snapshot
or release candidate.

It would be particularly useful if it could be extended for use by
developers as well as end users, by including the MinGW/MSys
development environment (gcc, make, flex, bison etc).

The last time I decided to put some effort into the native Windows
version, it took a good portion of 3 days to get to the point that I
could compile GRASS with the only problems being due to GRASS itself
rather than the development environment.

Many of the GRASS developers have access to a Windows system. If it
could be made so that getting to the point of being able to compile
GRASS on Windows took a few minutes rather than a few days, I think
that more of us would be willing to spend time on the native Windows
version.

One thing which I'm curious about: is it necessary to host all of the
packages on a common server? I know that the setup.ini file normally
only contains pathnames. Is that a restriction of setup.exe? Is it
something which would be easy to change? Would it be feasible to
configure the web server to use HTTP redirects to have non-OSGeo
packages (TIFF, JPEG, PNG, etc) retrieved from the original site?

-- 
Glynn Clements <glynn at gclements.plus.com>