------- Additional Comments From dsell at agleader dot com 2005-08-09 20:24 ------- We did come to the realization that we should be using doubles to store the values, because the values we were comparing were really too large for floats. However, the fact that rearranging the code or compiling on one platform vs another could cause different results seemed to be wrong behaivour.
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23298
