[GRASS5] The status of 5.0

[Markus Neteler]

On Fri, Mar 22, 2002 at 05:33:38PM -0600, Helena wrote:
> > While it would be nice to be able to perform thorough testing
> > "internally", I just don't see that happening.
> 
> We did quite systematic testing of a big part of GRASS5pre3 with Markus
> just recently. While there is absolutely no way that we could test every
> option for every kind of data combination we got a pretty good idea on how
> good (or bad) the code is.

Confirming that: Helena and me have tested the last months about 150 or
more modules in different situations (locations, projections, data sources
etc.). I think we identified more that 40 bugs of different severeness 
which are mostly fixed. Those not fixed are documented in the bug
tracker. Means, testing has happened a lot even if we didn't post that
to the list. Roger, if you look into the list archives you can see a lot
of bug mails I have sent from Helena's and my testing and their often
immediate fixes (most kudos to Glynn). More bugs we just fixed (see
CVS commit list archive:
http://grass.itc.it/pipermail/grass-commit/
Nov: 358, Dec 239, Jan 931, Feb 468, Mar 374 commits. Substract the 
release_branch and MAIN synchronization and you still have hundreds of
fixes.)
The situation is quite good comparing to earlier releases. Those bugs
which are found now, must be posted to the bug tracker. This system is
heavily below intended usage!

I fully agree qith you that we all hate unwanted surprises. But they do
not happy so often on average, probably only for certain modules.

Roger, in another mail you wrote that you had to heavily customize pre2.
Maybe it is a good idea to let us know about the changes. Probably
numerous are interesting for a wide audience and you gain a reduction of
workload in future code transitions.

> Vector support is the weekest and should be addressed by GRASS5.1 -
> it should be made clear for GRASS5.0 that vector support is limited and
> encourage
> others to contribute to its development. 
Yes, the 5.0 vector support will remain limited. It doesn't make sense
to work on oldish libraries when the 3D vector libs are "ante portas".

> Most of the rest really needs testing by a large users community as it is
> undoable by few people who just cannot have the variety
> of data required for such testing.

And: Most of the GRASS testing cannot be done automated. Andreas once
worked on the test-routines, however, to implement a rather complete 
test engine will be too costly.

[RT: ... more on this on Monday]

Markus