a5f9262adb
So toss a runtime error ("floating point error") when any of the math
library functions produce a non-finite value. Note that these are not
the only locations that can do that (simply dividing by zero will
produce an Inf), but it's still proper behavior.
[ Nothing here at this time. ]