I've modified all the trans math tests and PMC tests to check floating point operations by using a nearly-equal-to comparison macro rather than using perl to generate sprintf("%f"...) expected output.
If you're writing any new tests which check floaty things, can you follow the style of these tests as it will be much easier to unperl them in the future. Also in aid of reducing future work could all tests be written with CODE and OUTPUT sections. Alex Gough