[GRASS-dev] New WinGRASS Installer

[Moritz Lennert]

Marco,

On 12/07/08 10:23, Marco Pasetti wrote:
> Hi Moritz,
> 
>> In this image, the installer suggests that it will install a newer 
>> version of GRASS which is 6.3.0-4 instead of 6.3.0-3. In my 
>> understanding these are not two different versions of GRASS, but 
>> rather two different versions of the installer. If this is not 
>> correct, then it shouldn't be 6.3.0 in both cases. If it is correct, 
>> then why resinstall GRASS ?
> 

> yes, there's something to discuss here. You're right when you say that 
> it's not a different version of GRASS, it's still 6.3.0, but it's not 
> exactly a simple different version of the installer. Yes, the installer 
> may be different, but I usually release a new "installer" (that is a new 
> binary release) when I introduced changes in GRASS 

Which kinds of changes do you introduce in GRASS ?

> or its dependencies 
> binaries (adding new supports to GRASS or updating dependencies);

This is why it was suggested to separate GRASS from its dependencies in 
the installer...

But since (AFAIK) GRASS binaries compiled with one version of e.g. gdal 
or proj will not necessarily work with another version of these 
libraries, I see the problem of having to create new "versions" of GRASS 
whenever you update dependencies.

So, the question is double:

1) Should the winGRASS package of a particular version of GRASS be 
recompiled regularly with new versions of its dependencies ? Arguably, 
it would be much easier to decide on a particular version of these 
dependencies, in order not to introduce several possible sources of new 
bugs... And I don't think that having bleeding edge dependencies is 
absolutely necessary for winGRASS testing.
2) If yes, and you recompile regularly with new version of the 
dependencies, how should versioning be handled ?

> that 
> is because WinGRASS is still at a high development stage, and it may 
> happen that, between source releases, I add new features or complete 
> some To-Do items about supports/dependencies.

Well all depends on what you call "features". If any of this implies 
actual changes to the GRASS code, then you actually have a new version 
of GRASS.

> 
> This said, could we replace the word "version" in the message box with 
> "binary release" or simply "release"? that would be, actually  more 
> "honest"; do you agree?

I think we need to find an answer to above questions before we can 
decide. Personally, I would plead for a very stable build environment 
where only the GRASS source changes. AFAICT, your current winGRASS 
package is already more up to date and feature packed than most Linux 
distribution packages of GRASS, so no need to push this much further.

Moritz