TR1 Math Special Functions

[Benjamin Kosnik]

> Then, the next issue is that the various check_nan.cc tests must be 
> adapted to not rely on isnan being always available in std::tr1.

How do you want to proceed on this stuff?

I think it might make sense to check in with minor breakage and do the 
fixups afterwards. That way we can all work on it, and it will get done 
faster.

As it is, this patch is getting pretty ungainly.

We've got far to many of these types of patches at the moment...

-benjamin