greywolf@starwolf.starwolf.com asks:
> But since it's been fixed and is now quick as well, why "fix" it in such
> a horrific manner?

I could easily see what is "horrific" about code that requires several
programmers tens of minutes to decode, but I fail entirely to see what
is "horrific" about code which can be easily understood and maintained.

Which, by the way, on my 486 appears not to be noticably slower than
the so-called "quick" version:  in a quick test program, running 10,000,000
loops of ten tests each (unoptimized), the difference in runtime is about
1 second (I'd have to drop into single user mode to get good, repeatable
numbers) or about 10 nanoseconds each.  Which is very curious, since I don't
have a 100MHz clock.  I seriously doubt that I have run atan2() frequently
enough that the difference would have added up to as much time as people
spent tracking this down and fixing it -- and I have run a LOT of SPICE
jobs.