>
> > I do not particular like it. It outputs thousands of "No differences
> > encountered", that will scroll away any useful output. So,
> if any test
> > in between fails, there is no chance to see it unfortunately (there
> > seems to be no log by default). I just noticed, that one more test
> > failed - and it was only by accident.
>
> That's coming from your version of diff, not the test
> harness.  If you can
> find a portable way of suppressing it without removing the
> useful output, I
> will add it --- it may simply be a case of 2>/dev/null.
>

What about attached patch? (2>/dev/null does not work, diff outputs on
stdout).

Can we rely on the fact, that $? is set to exit code of command
substitution?

-andrej