@Jason S: your "fmod" variant will not work on a standards-compliant implementation. The C standard is explicit and clear (7.12.10.1, "the fmod functions"):
if y is nonzero, the result has the same sign as x
thus,
fmod(atan2(y,x)/M_PI*180,360)
is actually just a verbose rewriting of:
atan2(y,x)/M_PI*180
Your third suggestion, however, is spot on.