psql & regression (was: Error in new psql)

Peter Eisentraut wrote:

> On 1999-12-11, Jan Wieck mentioned:
>
> > I added a check to the rules regression test after that, to
> > ensure it never happens again. Unfortunately, Peter's
> > enforcement to use old psql for regression prevented it from
> > showing up.
>
> To be completely honest, I was just waiting to see what this was good
> for. As you have seen (or not), it was more or less disabled but still
> there.

Maybe it sounded the like, but I really did not wanted to
citicize your work. It was a great job and IMHO a big leap
forward in user friendliness of psql. I expect all this tab-
completion and help stuff to be highly appreceated and
honored. Let me be the first to explicitly say CONGRATS.

What I just wanted to point out is, that such a little,
subtle change in psql's input preprocessing could distort an
existing feature. In this case, it's totally clear to me
that is was only disabled and still there. But I only
stumbled over it because I tried to create a multi action
rule by hand to evaluate some comment I was writing on a
list. Without that, the proposed procedure (I outlined) to
update expected output would have broken the "rules"
regression test and stamped the broken results into expected.
So it probably wouldn't have been noticed until after
release.

And who can guarantee that this kind of flaw cannot happen
anywhere else? There are many, very old regression tests.
Some of them go back to the roots, Postgres 4.2, and I'm not
sure anyone ever looked at the expected results lately, if
they are really what SHOULD be expected. The tenk data for
example is something where even I don't know where it was
coming from, and I already joined the Postgres community with
release 4.2 back in 1994.

All this IMHO isn't really subject to your personal
responsibility. The interface of our interactive shell
needed the now happened polishing for some time. Instead I
wanted the backend developers to handle this major change in
psql, which is a core utility of the regression suite, not as
lax as past changes to it might have been. That's all.

Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#========================================= wieck(at)debis(dot)com (Jan Wieck) #