Parallel regress tests (was Re: FOREIGN KEY and shift/reduce)

wieck(at)debis(dot)com (Jan Wieck) writes:
> I'm thinking about another enhancement to the regression test
> now. Something where at least two sessions can run queries
> in a predefined order. Otherwise, something like the above
> cannot be checked during regression.

Yes, we could really use something like that.

> I'm not sure how that can be done with a standard shell, and
> that's a must. Maybe something using named pipes and so -
> will play around a little.

Would probably be more portable to write a little bit of C code that
opens several libpq connections and reads a script file with interleaved
commands to send to the different connections. OTOH, by the time you
got done, it might have much of psql in it (certainly at least the
display code).

Peter, in your slicing and dicing of psql, did you end up with anything
that might make this a feasible approach?

regards, tom lane