Jon Stockill writes:
>
> On Thu, 28 Aug 2003, Norman Vine wrote:
>
> > What errors are you getting from the Terra build ??
>
> After tweaking the Makefile to use the right compiler and flags, and point
> to glut running make gives:
>
> g++ -I/usr/X11R6/include -O2 -g -DSAFETY -DIOSTREAMH -c terra.cc
> In file included from Geom.h:24,
> from Heap.h:4,
> from GreedyInsert.h:4,
> from terra.h:4,
> from terra.cc:1:
> Vec2.h:43: ISO C++ forbids declaration of `ostream' with no type
> Vec2.h:43: `ostream' is neither function nor member function; cannot be
> declared friend
> Vec2.h:43: syntax error before `&' token
Jon
This is the top of my Terra/Vec2.h
#ifndef VEC2_INCLUDED // -*- C++ -*-
#define VEC2_INCLUDED
#include <iostream>
using std::ostream;
using std::istream;
This is a recnt change, perhaps you need to do a CVS up
If you have this change then AFAIK you can probably work around
this by explicitly calling the fully decorated object
i.e. line 43 Vec2.h
// Input/Output methods
friend std::ostream& operator<<(std::ostream&, const Vec2&);
friend std::istream& operator>>(std::istream&, Vec2&);
> > AFAICT we want to use the Python script which if called with no arguments
> > should print some help i.e.
> >
> > $ python src/Prep/TerraFit/terrafit.py
>
> AIUI this is just a script which drives terra though, not a standalone
> solution?
Correct but you can't call terra directly as the data needs to be massaged
both on input and output
> > Usage: src/Prep/TerraFit/terrafit.py
> > -h | --help
> > -m | --minnodes 50
> > -x | --maxnodes 600
> > -e | --maxerror 50
> > -f | --factor 0.033333
> > -v | --version
> > [file] | [path to walk]
> >
> > Algorithm will produce at least 50 fitted nodes, but no
> > more than 600. Within that range, the algorithm will stop
> > if the maximum elevation error for any remaining point
> > drops below 50 meters.
>
> Are these sensible values, or should we be pushing maxnodes up, and
> maxerror down (and if so, how far) - I'm guessing that fitting a better
> surface will take longer, I'm not particularly worried about this, as long
> as we get a good result for the finished scenery - also more points=more
> triangles, and we need to ensure we have scenery that's going to be usable
> on a wide range of hardware.
< aside >
Ah now we are leaving science and getting into art :-)
< /aside >
The factor value is the only one which is required to be as above
< it sets the scaling for vertical scale vs horizontal scale of the input data >
AFAIK these values are the ones that Curt was using as I just copied them
from his 'C' driver program for the arrayfit process
FWIW
I would decrease the maxerror considerably < perhaps to 2 at least to 5> as
AFAICT Terra will always insert nodes ordered by 'max error' and will always
quit when maxnodes is reached.
As to what to use for maxnodes, I am not sure.
I wouldn't worry to much about run time as Terra is 'remarkably' quick.
i.e using Terra even when driven with a Python script should be substantially
more then an order of magnitude quicker then using arrayfit
It's worth getting Terra running before starting the global decimation :-)
HTH
Norman
_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel
