LongBitsToDouble Called With A Narrow Argument

Reports Double.longBitsToDouble(..) calls whose single argument is provably narrower than long: an int or char literal, or a name declared int, short, byte or char.

Double.longBitsToDouble reinterprets the 64 bits of its argument as an IEEE-754 double. A narrower argument is first widened to long the ordinary Java way, sign-extending the value — which changes the bit pattern the method actually reads, most visibly by corrupting the sign and exponent bits:

The call compiles without a warning, since int widens to long implicitly, so the mistake is silent: the method returns a value, just not the one the bit pattern was meant to produce.

Give the method a value that is already long, either by declaring the source value long or by reconstructing the bit pattern with a long literal: