Um, I hate to break this to you folks, but, barring an AI diff being
available, the regression tests might need to be adjusted anyway, since
they rely on a particular table output format which I cannot guarantee to
(or even want to) reproduce exactly.
But I hear the tests were going to be revised for 7.0 anyway, right? ;)
Since I got you into this mess I would certainly try to help to resolve
it, but that's a future issue.
-Peter
On Oct 11, Peter Eisentraut mentioned:
> On Oct 11, Thomas Lockhart mentioned:
>
> > > some query;
> > > *** comment
> > > *** comment
> > > \p
> > > \r
> > > more queries;
> > > which should probably be changed anyway to something like
> > > -- comment
> > > -- comment
> >
> > Actually, this is probably testing that the buffer reset actually
> > clears the lines, which wouldn't be as obvious if there were only
> > legal SQL preceeding it. Maybe leave it as-is??
>
> I think I figured that out: the *** comments actually show up in the
> output of the regression tests as an aid to a person glancing at the
> results. If the regression tests wanted to test psql then they could
> certainly do a lot more fun things than that.
>
> I was planning on implementing an \echo command which could easily be
> dropped in there as a more elegant solution.
>
> Which makes me think. The server regression tests should certainly not
> rely on some particular psql functionality, just as a matter of principle.
> But if I'm ever really bored I could write a separate psql regression
> test. No promise though.
>
> -Peter
>
>
--
Peter Eisentraut Sernanders vaeg 10:115
peter_e@gmx.net 75262 Uppsala
http://yi.org/peter-e/ Sweden
