InstantDeserializer deserializes the nanosecond portion of fractional timestamps incorrectly: -1.000000001 deserializes to 1969-12-31T23:59:59.000000001Z instead of 1969-12-31T23:59:58.999999999Z

[Strongbeard]

Version

2.15.4 - 2.19 and the current master branch (commit c84b1a9 as of this post).
I have not tested anything earlier than what I plan to use.

Description

Please correct me if I am misguided on how negative fractional representations of timestamps should be parsed into java.time.Instant objects. I expected -1.000000001 to deserialize to 1969-12-31T23:59:58.999999999Z, but currently it deserializes to 1969-12-31T23:59:59.000000001Z.

I assume when "splitting" the fraction into its whole seconds and nanoseconds portions, the negative sign should be distributed to both. In other words: -1.000000001 == -1 + -0.000000001, but instead the behavior appears to be -1.000000001 == -1 + 0.000000001.

Strangely, any scenario where the seconds portion is 0 deserializes as I expect it to. For example, -0.000000001 deserializes to 1969-12-31T23:59:59.999999999Z. For some reason, the distribution of the negativity from seconds to nanoseconds is preserved here.

Adding the following unit test to datetime/src/test/java/tools/jackson/datatype/jsr310/deser/InstantDeserTest.java should demonstrate the issue, where test 4 fails and test 5 succeeds:

@Test
public void testDeserializationAsFloat04() throws Exception {
    Instant actual = READER.readValue("-1.000000001");
    Instant expected = Instant.ofEpochSecond(-1L, -1L);
    assertEquals(expected, actual);
}

@Test
public void testDeserializationAsFloat05() throws Exception {
    Instant actual = READER.readValue("-0.000000001");
    Instant expected = Instant.ofEpochSecond(0L, -1L);
    assertEquals(expected, actual);
}

The following error is produced when executing the tests:

[ERROR] tools.jackson.datatype.jsr310.deser.InstantDeserTest.testDeserializationAsFloat04 -- Time elapsed: 0.128 s <<< FAILURE!
org.opentest4j.AssertionFailedError: expected: <1969-12-31T23:59:58.999999999Z> but was: <1969-12-31T23:59:59.000000001Z>
        at [email protected]/org.junit.jupiter.api.AssertionFailureBuilder.build(AssertionFailureBuilder.java:151)
        at [email protected]/org.junit.jupiter.api.AssertionFailureBuilder.buildAndThrow(AssertionFailureBuilder.java:132)
        at [email protected]/org.junit.jupiter.api.AssertEquals.failNotEqual(AssertEquals.java:197)
        at [email protected]/org.junit.jupiter.api.AssertEquals.assertEquals(AssertEquals.java:182)
        at [email protected]/org.junit.jupiter.api.AssertEquals.assertEquals(AssertEquals.java:177)
        at [email protected]/org.junit.jupiter.api.Assertions.assertEquals(Assertions.java:1145)
        at tools.jackson.datatype.javatime/tools.jackson.datatype.jsr310.deser.InstantDeserTest.testDeserializationAsFloat04(InstantDeserTest.java:121)
        at java.base/java.lang.reflect.Method.invoke(Method.java:580)
        at java.base/java.util.ArrayList.forEach(ArrayList.java:1596)
        at java.base/java.util.ArrayList.forEach(ArrayList.java:1596)

If my assumptions are correct, then the fix appears more complicated than just changing the negativeAdjustment boolean passed to the call to DecimalUtils.extractSecondsAndNanos(BigDecimal, BiFunction, boolean) within the _fromDecimal(DeserializationContext, BigDecimal) function of InstantDeserializer to false since that causes unit test ZonedDateTimeSerTest#testSerializationAsTimestamp01NegativeSeconds to fail.

Related Issues

I believe this is related to the implementation of the fix for issues #304, #69, and #132.