Scas might misassemble the kernel's floating point routines

[pixelherodev]

During 1.0 testing for scas, I discovered that there was a hash mismatch in the produced page 2 binary for the kernel (pages zero and one match 100% the output of sass :). There were a few bugs lurking, including a bug in ifndef, most of which have been resolved as of 1.0-rc4 (hence the new release candidate - this fixes some serious issues). The only remaining issue, according to a diff of the hexdump, lies in the floating point routines.

Given the nature of the recent issues, I'm fairly certain it's either an issue with conditional assembly or with macro expansion. Both of these are more fragile than they ought to be, unfortunately. This is absolutely a release blocker of 1.0.