[GRASS5] r.fill.dir segfaults

[H Bowman]

> > I was able to reproduce the bug:
[...]
> > r.mapcalc "elev.float = elevation.dem * 1.0"
> > r.fill.dir in=elev.float elev=blah dir=dir
[...]
> This works fine for me (at least, it doesn't segfault), with both
> 5.0.2 and a recent build from HEAD.


segfaults for me for both a DCELL and CELL maps on 5.0.2.
 (same map, second is round(x) the first)

The integer map (output below) didn't segfault until the second run of
the program.


note: the module doesn't work correctly on maps with negative elevation,
had to shift so entire map was positive.


Hamish


====================================================
GRASS:~ > r.fill.dir in=bf elev=bfo dir=BFdirn
Reading map...
Filling sinks...
Determining flow directions for ambiguous cases...
Downward pass 1
Upward pass 1

Downward pass 2
Found 547 unresolved areas
wtrshed pass 1
wtrshed pass 2
wtrshed pass 3
wtrshed pass 4
wtrshed pass 5
wtrshed pass 6

Repeat to get the final directions...
Downward pass 1
Upward pass 1

Downward pass 2
Upward pass 2
Found 84 unresolved areas

GRASS:~ > r.fill.dir in=bfo elev=bfo2 dir=BFdirn
Reading map...
Filling sinks...
Determining flow directions for ambiguous cases...
Downward pass 1
Upward pass 1

Downward pass 2
Upward pass 2
Found 84 unresolved areas
wtrshed pass 1
wtrshed pass 2
wtrshed pass 3
wtrshed pass 4
wtrshed pass 5
wtrshed pass 6
wtrshed pass 7
Segmentation fault