BigDecmial generation fails on Java 11

[iljakorneckis]

When generating BigDecimals on Java 11, JavaFixture populates the stringCache field with a random not-a-number string. This affects the .toString() method at the very least and, by proxy, affects JSON generated by ObjectMapper. The .intValue() also produces zero. On Java 17, the same field is populated by a correct string representation of the decimal.

Could it be that the reason it works on Java17 is the module system preventing access to internals and forcing JF to use a constructor?

Test expected to pass

@Test
void bigDecimalTest() {
     var bigDecimal = Fixture.fixture().create(BigDecimal.class);

     assertThat(bigDecimal.intValue()).isNotZero();
     assertThat(new BigDecimal(bigDecimal.toString())).isEqualTo(bigDecimal);
}

Json generated by Object mapper:

The cpc field is a BigDecimal. It's not even valid Json, which is more of a problem with ObjectMapper than JavaFixture tho.

{
 "cpc":EMxJJRhq3XBJVv8vayJ1ivNOzWoO6tuegfY1DBCMLb9MrfYKdJSq19RhKB3COzqEoEOmfsD8naQbKvzTXLFyKmzYu45wV6YiDbTt2X4OimoFZ3hk7fUm4zmdOYK7TuOC,
 "category_id":-1313760193,
}

Java 11 BigDecimal

Java 17 BigDecimal

[akutschera]

Yes. That is a bug. We should not rely on reflection for a BigDecimal, similar to a BigInteger.
Therefore I propose we add it to the SpecialSpecimen classes and create one using a random value in a constructor.

[Nylle]

Fixed in version 2.9.5