[GRASS-dev] Wingrass and TclTk
michael.barton at asu.edu
Wed Oct 31 20:55:53 CET 2007
I'd argue for something between these positions.
I'd stick to just a GRASS installation to keep it simple and relatively
small. But, I'd really recommend that there is the option of a SINGLE
installation package for users. Most of the Windows users I deal with are
intimidated by a package that they download and then find out that they have
to go somewhere else and install something different just to use the first
So can we include TclTk and Mysys, if needed, in a single package that
allows all of GRASS to run?
On 10/31/07 8:25 AM, "Moritz Lennert" <mlennert at club.worldonline.be> wrote:
> On 31/10/07 13:58, Benjamin Ducke wrote:
>> Glynn Clements wrote:
>>> Moritz Lennert wrote:
>>>> So, I still believe that:
>>>> rem Path to the shell command
>>>> rem set GRASS_SH=c:\msys\1.0\bin\sh.exe
>>>> is a better solution (since allowing the user to install what they want
>>>> where they want) than
>> Yes, but often (especially my type of user), they don't know enough
>> to want anything (see below).
>> Anyway, a knowledgeable user can always adjust those vars by hand!
>>> Agreed. The above is the only location where a Bourne shell is likely
>>> to be found. If it's anywhere else, the user will have to set it
>> Or we supply it as part of the WinGRASS binary distribution! (see below)
>>> The relative path will only work if grass63.bat is installed in
>>> c:\msys\1.0, so there's no benefit to using it.
>> Not quite. With the setup above, GRASS can be installed _anywhere_ on
>> the file system. The only prerequisite is that grass63.bat sits
>> in the same folder as the GRASS install dir.
>> E.g. on my harddisk I have grass63.bat in c:\WinGRASS\
>> and the GRASS dir (grass-6.3.cvs) in the same folder.
>> I then copied C:\msys\1.0\bin, dll and share
>> (at least the parts that I needed for using the shell) into
>> This way, I have a completely self-contained GRASS distribution.
>> I can add more bits an pieces (such as R) easily as needed and
>> at the end, put everything into on ZIP for distribution.
>> I don't know what could be easier. Especially since this approach
>> does not interfere with an MSYS already installed on the user's
> But it might install the same programs twice on the machine and probably
> bloat the grass distribution by quite a lot (don't forget that most
> people do not have broadband) ...
> IMO, we should stick to a basic wingrass distribution to which people
> can add what they want in terms of other programs. If they want access
> to R, they can add R's bin + lib directories to the path and that's it.
> If you have a particular audience that needs a series of add-ons and
> tweaks than you can provide a special package for them, but I prefer to
> distribute a simple package without any unnecessary additions (and I
> consider msys an unnecessary addition in this context as there is hardly
> anything you cannot do without scripts - it just might not be as
> convienient). Don't assume that everyone has the same needs as you.
> And up to now feedback has been that installation of the wingrass
> package is very easy...
> However, I agree that we could maybe provide more detailled
> documentation on how to integrate different packages (at least msys).
>> The benefit for the user is that only one .bat is visible in
>> the top dir: it's clear what to click on and there is no
>> searching in bin grass-6.3.cvs/bin or wherever to find
>> a startup-script.
> If the instructions in the readme are not clear enough, please provide
> suggestions for improvement. Up to now I have not received any feedback
> from users not being able to find how to launch grass.
>> Please keep in mind, that my Windows target users will often not
>> have the ability (or patience) to install MSYS by themselves and set
>> the appropriate vars in some obscure batch script -- they just
>> want to click and run GRASS!
> And so they _can_ with the current distribution, except for the fact
> that they have to install ActiveTcl if they want the gui (and the fact
> that the instructions need a reference to how to start grass in text
> mode). This should be solved by giving the option of downloading a grass
> package including a free tcl/tk installation.
> My point above concerning msys being unnecessary is obviously arguable
> and if there is a general opinion that we should include it, we can (but
> without any need to change the grass63.bat except for uncommenting the
> line about where to find sh.exe and another to set the necessary PATH).
Michael Barton, Professor of Anthropology
Director of Graduate Studies
School of Human Evolution & Social Change
Center for Social Dynamics & Complexity
Arizona State University
More information about the grass-dev