GRASS: Puzzle with r.cost - is more programming the solution?

[James Steele]

Dear colleagues,

One of my research students is using r.cost, and finds that when the module
grows a cost surface radially from the origin cell across a cost surface
with cells of a uniform value (e.g. 1), the resultant contours of increasing
cost are less-than-ideal approximations of the expected pattern (which would
be a nested series of perfect circles incrementing in radii of 1 unit of
cost). Specifically, the output has a kind of octagonal pattern, and this is
something we need to fix. And yep, she is using the Knight's Tour switch.

Before we start messing with the source code ourselves, can I ask for any
advice from any folks who may have troubleshot this problem already? Are we
being dolts?

Many thanks in advance.

Regards,
James Steele

-------------------------------------------------------
James Steele, Ph.D., Lecturer,
Dept of Archaeology,
University of Southampton,
Highfield,
Southampton SO17 1BJ,
England.

E-Mail: tjms at soton.ac.uk
Telephone: (01703) 594198
Fax: (01703) 593032
=======================================================