BigDecimal's double can lose precision in surprising ways.

  // these are the same:
  new BigDecimal(0.1)
  new BigDecimal("0.1000000000000000055511151231257827021181583404541015625")

Prefer the BigDecimal.valueOf(double) method or the new BigDecimal(String) constructor.

NOTE BigDecimal.valueOf(double) does not suffer from the same problem; it is equivalent to new BigDecimal(Double.valueOf(double)), and while 0.1 is not exactly representable, Double.valueOf(0.1) yields "0.1". As long as FloatingPointLiteralPrecision doesn't generate a warning, BigDecimal.valueOf is safe.