Float.ceil returns incorrect result

[mtokioka]

Float.ceil(0.0) returns different value between elixir 1.3.2 and 1.4.0-rc.1

Interactive Elixir (1.3.2) - press Ctrl+C to exit (type h() ENTER for help)
iex(1)> Float.ceil(0.0)
0.0

Interactive Elixir (1.4.0-rc.1) - press Ctrl+C to exit (type h() ENTER for help)
iex(1)> Float.ceil(0.0)
1.0

[ardhitama]

Failing test in float_tests.exs

test "ceil/1" do
   assert Float.ceil(12.524235) === 13.0
   assert Float.ceil(-12.5) === -12.0
   assert Float.ceil(-12.524235) === -12.0
   assert Float.ceil(7.5e3) === 7500.0
   assert Float.ceil(7.5432e3) === 7544.0
   assert Float.ceil(7.5e-3) === 1.0
   assert Float.ceil(-12.32453e4) === -123245.0
   assert Float.ceil(-12.32453e-10) === 0.0
   assert Float.ceil(0.32453e-10) === 1.0
   assert Float.ceil(-0.32453e-10) === 0.0
   assert Float.ceil(1.32453e-10) === 1.0

   assert Float.ceil(0.0) === 0.0
 end

 test "ceil/2 with precision" do
   assert Float.ceil(12.524235, 0) === 13.0
   assert Float.ceil(-12.524235, 0) === -12.0

   assert Float.ceil(12.52, 2) === 12.52
   assert Float.ceil(-12.52, 2) === -12.51

   assert Float.ceil(12.524235, 2) === 12.53
   assert Float.ceil(-12.524235, 3) === -12.524

   assert Float.ceil(12.32453e-20, 2) === 0.01
   assert Float.ceil(-12.32453e-20, 2) === 0.0

   assert Float.ceil(0.0, 0) === 0.0
   assert Float.ceil(0.0, 2) === 0.0
   assert Float.ceil(0.0, 3) === 0.0
 end

[tsubery]

Zero as a floating point is encoded in a special form, without a biased exponent.
As far as I can tell, this line assume the number has a biased exponent and leads the code to line 207 instead of 203.