[GRASS-dev] r39128 automatically choose appropriate geometry type

[Hamish]

Helena wrote:
> where do the current somewhat broken items in nviz belong in your list,
> I am talking about
>
> 1. fixing sites with attributes handling (I think Maris
> has been looking into it but I am not sure how difficult it is
> to fix it)
>
> 2. nviz crashing or freezing with reclassed raster (it
> should at least give an error message and exit) - there was some
> discussion about it - 64bit issue? but I am not sure whether it was
> fixed
>
> 3. and a minor but annoying thing that nviz won't start
> from wxGUI (under File>NVIZ) in winGRASS
> Should we consider things that worked in 63 but don't work
> now critical?
> Even if only few people may use them?


the term "critical" might be a bit hyperbolic, but I think the general
idea is just to go through an extra mental step and ask yourself if the
change is really a big deal or not, and coupled to that how much of a
dent it could cause to other things if it contained some unintended
consequence. A minor tcl edit probably isn't going to cause stability
problems to the wider system for example, but as we saw with #654 in
rc5 a seemingly innocent change to libgis can.

To me, File->NVIZ not working in the WinGrass wxGui is a fairly major
bug. New users don't have (obvious) access to one of our flagship
components. The cause of one is still a complete mystery to me.
Also, there seems to be something more needed than just PATHEXT for
v.type and g.gui wrappers.


To me, nviz crashing on reclass maps is a medium priority bug; but I
wouldn't stall 6.4.0 for it. For the record it does seem to exit with
an error, even if that message isn't all that helpful.

Sites with attributes is a regression, but the positive side of that
is that we can checkout by date and narrow it down to the exact commit
that broke it (someone has already done that?) so not a complete mystery.

If clear bugfixes are available I don't really object to backporting
them, it's just trying to avoid the trap of rushing in all the unfinished
business before the deadline, which leads to untested code & a poor
release. The other thing to watch out for is cosmetic edits to strings
just before release. All 10-20 translations then need to be redone but
it's too late...


2c more,
Hamish