DateTime.jd ignores fractional seconds

[jeremyevans]

Environment Information

Happens with both:

jruby 9.4.2.0 (3.1.0) 2023-03-08 90d2913fda OpenJDK 64-Bit Server VM 20.0.1+9-29 on 20.0.1+9-29 +jit [x86_64-mswin32]
jruby 9.4.5.0 (3.1.4) 2023-11-02 1abae2700f OpenJDK 64-Bit Server VM 17.0.9+9-1 on 17.0.9+9-1 +jit [x86_64-OpenBSD]

Expected Behavior

Respects fractional seconds, as in CRuby:

$ ruby -r date -e "p DateTime.jd(2454984, 15, 57, 3/2r).sec_fraction"
(1/2)

Actual Behavior

Ignores fractional seconds:

$ jruby -r date -e "p DateTime.jd(2454984, 15, 57, 3/2r).sec_fraction"
(0/1)

Found by Sequel's tests after a refactor.

[headius]

I wonder if this is an out of date stdlib.

[headius]

Welp, there's yer problem right there:

jruby/core/src/main/java/org/jruby/ext/date/RubyDate.java

Lines 497 to 508 in 9f38144

	public static IRubyObject _valid_time_p(ThreadContext context, IRubyObject self,
	IRubyObject h, IRubyObject m, IRubyObject s) {
	long hour = normIntValue(h, 24);
	long min = normIntValue(m, 60);
	long sec = normIntValue(s, 60);
	if (valid_time_p(hour, min, sec)) {
	return timeToDayFraction(context, (int) hour, (int) min, (int) sec);
	}
	return context.nil;
	}

We normalize seconds into a long, so there's no where for a rational to go. This logic is just out of date since CRuby started accepting Rational here.

[headius]

The CRuby code starts here and involves many macros and local variable twiddling... so I did not start a port attempt. It's clear this logic has grown and we'll need to expand our handling of the time elements accordingly.

https://github.com/ruby/ruby/blob/7ffee5681f85de3fe74c25a90e05e31616113123/ext/date/date_core.c#L7659-L7717