Fix incorrect exception when exponent is fractional for Infinity base (#453)

* Handle Infinity base in power’s precision estimator

* Apply precision path only when x > 1 && x.finite?

Author: troy-dunamu

lib/bigdecimal.rb

@@ -194,7 +194,7 @@ def power(y, prec = 0)
       end
       ans.mult(1, result_prec)
     else
-      if x > 1
+      if x > 1 && x.finite?
         # To calculate exp(z, prec), z needs prec+max(z.exponent, 0) precision if z > 0.
         # Estimate (y*log(x)).exponent
         logx_exponent = x < 2 ? (x - 1).exponent : Math.log10(x.exponent).round

test/bigdecimal/test_bigdecimal.rb

@@ -1851,10 +1851,16 @@ def test_power_of_positive_infinity
     assert_positive_infinite_calculation { BigDecimal::INFINITY ** 1.quo(1) }
     assert_positive_infinite_calculation { BigDecimal::INFINITY ** 1.0 }
     assert_positive_infinite_calculation { BigDecimal::INFINITY ** BigDecimal(1) }
+    assert_positive_infinite_calculation { BigDecimal::INFINITY ** 1.quo(2) }
+    assert_positive_infinite_calculation { BigDecimal::INFINITY ** 0.5 }
+    assert_positive_infinite_calculation { BigDecimal::INFINITY ** BigDecimal(0.5) }
     assert_equal(1, BigDecimal::INFINITY ** 0)
     assert_equal(1, BigDecimal::INFINITY ** 0.quo(1))
     assert_equal(1, BigDecimal::INFINITY ** 0.0)
     assert_equal(1, BigDecimal::INFINITY ** BigDecimal(0))
+    assert_positive_zero(BigDecimal::INFINITY ** -1.quo(2))
+    assert_positive_zero(BigDecimal::INFINITY ** -0.5)
+    assert_positive_zero(BigDecimal::INFINITY ** BigDecimal(-0.5))
     assert_positive_zero(BigDecimal::INFINITY ** -1)
     assert_positive_zero(BigDecimal::INFINITY ** -1.quo(1))
     assert_positive_zero(BigDecimal::INFINITY ** -1.0)