Found the bug with my little eye. It was quite quick to locate in the
end. Shows that it does pay not to rely on these debugging tools too
heavily.
It was in some code I'd lifted from CENTROID's KPG1_CENSG by ASTman aka
Lancs_lad. (-: All the necessary evidence was in a mail sent to David
before trying valgrind, that included the snippet of code. I guess it
doesn't lead to disaster in CENTROID.
I'm somewhat dismayed that there's no valgrind analysis that indicates
bounds mismatches in subroutine calls.
Thanks chaps for the assistance.
Malcolm
|