I've committed fixes for the recently reported problem with adding
timestamp with time zone to intervals. As pointed out earlier, there
were one or two places which did not make the right choice between
"zoneful" and "zoneless" forms of support routines.
I've also committed some additions to the horology regression test to
exercise more possible input formats of date/time types. These should be
expanded further after 7.2 is released.
Regression tests pass, and Tom Lane has volunteered to fix up the
alternate horology regression outputs.
- Thomas
