pg_regress gripes

A couple glitches to work out in pg_regress:

1. 'make installcheck' fails if PGPORT isn't set in environment. Should
omit '-p $PGPORT' options from program invocations unless PGPORT is set.
Actually, I think you should omit -p period; the programs are perfectly
capable of reading PGPORT from the environment for themselves.

2. I get this behavior:

test hash_index ... ok
test misc ... FAILED
test select_views ... ok
test alter_table ... ok
test portals_p2 ... ok
test rules ... ok
test foreign_key ... ok
test limit ... ok
test plpgsql ... ok
test temp ... ok

=================================================
0 of 75 tests passed, 0 failed test(s) ignored.
=================================================

(The misc failure is from a bug in some other stuff I'm working on;
I show it just to illustrate that pg_regress's helpful totals are
completely incorrect.) I presume the output is more useful for you,
which probably means that echo behaves differently on your platform.
Possibly should be looking for '\.\.\. *ok' and so on?

regards, tom lane