The failure at: http://tinyurl.com/y27tlc looks like problem with the test itself. Anybody's in position to fix it? I think the same operator<< was causing problems for other compilers and John has a patch, but I have no idea if this patch will apply to cw-9.4. If we exclude stale (Nov 22) test failures that are fixed since, those borland, 6.5, 7.0 and cw-9.4 are four remaining failures of the math library, it would be great to see them go away. Thanks, Volodya