[Gzz] Re: Problems with coordinate systems

[Tuomas Lukka]

On Sat, Feb 08, 2003 at 05:46:06PM +0200, Matti Katila wrote:
> On Sat, 8 Feb 2003, Tuomas Lukka wrote:
> > On Sat, Feb 08, 2003 at 05:19:21AM -0500, Matti Katila wrote:
> >>  public final class GLVobCoorder extends AffineVobCoorder {
> >> -public static final String rcsid = "$Id: GLVobCoorder.java,v 1.67 
> >> 2003/02/06 11:53:14 tjl Exp $";
> >> +public static final String rcsid = "$Id: GLVobCoorder.java,v 1.68 
> >> 2003/02/08 10:19:20 mudyc Exp $";
> >>      public static boolean dbg = false;
> >>      private static void pa(String s) { System.err.println(s); }
> >>  
> >> -    float[] floats = new float[10000];
> >> +    float[] floats = new float[10000*2];
> >>      int nfloats = 0;
> >>  
> >> -    int[] inds = new int[5000];
> >> +    int[] inds = new int[5000*2];
> >>      int ninds = 1; // zero is special (the root)
> > 
> > Is there a reason for 10000*2 and not 20000? Please try to keep the code 
> > simple.
> 
> Yes there is. It should say that there is a problem and the fix is dirty.
> 
> If anyone want to *use* pp you need more coordinate systems to fit in 
> memory or it crashes. Coordinate systems are very light weight to create 
> but the summ of them all seems to expand quite heavily because buoy is a 
> view of another view and there can be a lot of coordinate systems.

Yes. There is a PEG about fixing the cause, vobscene_recursion.

Also, of course it shouldn't crash, it should expand the arrays.
But it would be nicest to catch the indexOutOfBoundsException *outside*
the view...

        Tuomas