While doing the regression tests for class Time (RubyUnit is
a wonderful thing!), I came across an odd problem:

	p Time.local(2001, 4, 1).to_s

Results in:

	Sat Mar 31 23:00:00 EST 2001

This happens with all years from 1988 on, with a TZ of CST or EST.

The problem seems to be in time.c:make_time_t(tptr, fn)
Line 306:
    guess += (tptr->tm_mday - tm->tm_mday) * 3600 * 24;

Not all days have 24 hours!  In particular, days affected by Daylight
Savings Time (in the US, that's usually the first Sunday in April and
last Sunday in October) will have 23 or 25 hours.

Might it be simpler to use the platform's mktime() function to
avoid these issues? (I believe that's a Posix-ism).

Thanks,

/\ndy