DST bug in second form of Time.local

Interesting

Note that I was referring to the hour difference, not the GMT offset vs string. However, as of revision 4842, it looks like you've switched to GMT offset. Anyway, I still see the bug.

JRuby's RubyTime just ignores the last 4 arguments altogether:

in createTime():

if (args.length == 10) {
args = new IRubyObject[] { args[5], args[4], args[3], args[2], args[1], args[0], runtime.getNil() };
}

So, no matter what is specified in the last 4 arguments, they don't affect the end result of Time.local invocation.

This also leads to one of activesupport unit test failures:

4) Failure:
test_seconds_since_midnight_at_daylight_savings_time_end(TimeExtCalculationsTest)
[/opt/work/activesupport/./test/core_ext/time_ext_test.rb:35:in `test_seconds_since_midnight_at_daylight_savings_time_end'
/opt/work/activesupport/./test/core_ext/time_ext_test.rb:27:in `with_timezone'
/opt/work/activesupport/./test/core_ext/time_ext_test.rb:27:in `test_seconds_since_midnight_at_daylight_savings_time_end'
/opt/work/activesupport/./test/core_ext/time_ext_test.rb:27:in `run']:
before DST end.
<5400> expected but was
<9000.0>.

The test in the activesupport test suite uses the long form of Time.local to set DST flag, and fails, since the flag is ignored.

E.g., with TZ=US/Eastern:

MRI:
Time.local(0,30,1,30,10,2005,0,0,false,ENV['TZ']).dst?
Time.local(0,30,1,30,10,2005,0,0,true,ENV['TZ']).dst?

returns false, true.

JRuby both times returns false.

Punting issues from 1.1 RC2 to 1.1 final.

A patch to clean up time creation a bit...it doesn't pass marshalling though, and I haven't time to fix it better.

Move to post 1.1, since the time stuff is a bit involved at the moment. Good 1.1.1 fix if we get it done.

Still broken going into 1.2...here's updated output:

[headius @ cnutter:~/projects/jruby]
$ jruby -e "p Time.local(1,2,3,27,2,2001,'Tue',287,true,'MST')"
Tue Feb 27 02:02:01 -0700 2001
[headius @ cnutter:~/projects/jruby]
$ ruby -e "p Time.local(1,2,3,27,2,2001,'Tue',287,true,'MST')"
Tue Feb 27 02:02:01 -0600 2001

On second inspection it appears that JRuby's output is actually more correct than MRI's...MST is always supposed to be -0700, not ever -0600. So I'm going to knock this one off as resolved since it seems like we're talking solely about a display issue that's probably fuzzy anyway. If there's a real issue here, like that we're not accurately representing time internally, open a separate issue.