A DESCRIPTION OF THE PROBLEM :
The internal method java.math.BigDecimal.fractionOnly() erroneously returns true for large scale values because it calculates `int - int` which can overflow; instead it should first cast one operand to `long`.

This leads to an incorrect exception message for java.math.BigDecimal.longValueExact() which claims "Rounding necessary" while the message should be "Overflow".

IMPORTANT: If you fix this numeric overflow in BigDecimal.fractionOnly(), you must also fix it here: https://github.com/openjdk/jdk/blob/d8158897c3d0dbea46e4f55ad8b501252d88b7e1/src/java.base/share/classes/java/math/BigDecimal.java#L3628 (same numeric overflow problem for large `scale` values)
Otherwise this might circumvent the fix for JDK-8211936.


STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Run the following two statements and inspect the exception messages:
new BigDecimal("1e" + (Integer.MAX_VALUE - 1)).longValueExact()
new BigDecimal("1e" + Integer.MAX_VALUE).longValueExact()


EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
java.lang.ArithmeticException: Overflow
java.lang.ArithmeticException: Overflow

ACTUAL -
java.lang.ArithmeticException: Overflow
java.lang.ArithmeticException: Rounding necessary