r.le program bugs

BAKERWL BAKERWL at corral.uwyo.edu
Mon Jun 28 14:13:51 EDT 1993

Those of you who have downloaded Version 2.0 of the r.le programs from
the ftp server "incoming" directory should be aware of one potentially
significant bug that is not yet fixed and another that is.

The first, more serious bug is in r.le.patch and r.le.dist.  When two
polygons intersect along diagonal pixels (bishops pattern) the trace.c
routine does not determine their areas correctly, as the app_island()
routine fails to subtract the area of the island.  What this means is
that if there are many intersecting polygons the reported values for
r.le.patch attribute (e.g., A1-4), size, shape, and edge may be off
by an unknown amount.  I found an error of 3 pixels for a polygon that
was 350 pixels in size, but larger errors are possible.  The r.le.dist
program may not be affected-I have not been able to determine this yet,
but r.le.dist uses the same tracing routine.

Unfortunately, the algorithmic solution to this problem is non-trivial.
I will be working on it, but also will be gone much of the summer.  The
r.le.tex program is not affected.

The second bug is also in the tracing routines in r.le.patch and r.le.dist.
These programs would provide a message such as "G-calloc failure-out of
memory" when the map to be analyzed contains more than about 400 patches.
This memory bug has been fixed, so that r.le.patch now can handle 1500+
patches (on my machine with 16 Mb of memory).  I will post the bug fix,
which is relatively simple on about July 9.

Please continue to inform me of other bugs you find in the r.le programs.

	W. Baker
	bakerwl at corral.uwyo.edu
	Univ. of Wyoming, Dept. of Geography

More information about the grass-user mailing list