> I was intending to suggest that much of what Robins has submitted
> doesn't belong in the core regression tests, but could usefully be put
> into an optional set of "big" regression tests. We already have a
> "numeric_big" test in that spirit. What seems to be lacking is an
> organizational principle for this (should the "big" tests live with the
> regular ones, or in a separate directory?) and a convenient "make"
> target for running the big ones. Once we had a setup like that, we
> could get some or all of the buildfarm machines to run the "big" tests,
> but we wouldn't be penalizing development work.
I have been suggesting something upon that line in some of the reviews
I've posted about Robins non regression tests, if they were to be rejected
on the basis that they add a few seconds for checks. They are well made to
test corner cases quite systematically, and I feel that it would be sad if
they were lost.
Looking at the GNUmakefile, ISTM that there could be a
{parallel,serial}_big_schedule derived from {parallel,serial}_schedule by
appending a few things in a separate file:
parallel_big_schedule: parallel_schedule parallel_big cat $^ > $@
bigcheck: ... parallel_big_schedule ... --schedule=$(srcdir)/parallel_big_schedule ...
and idem for serial & bigtest.
Also, the serial_schedule should be automatically derived from the
parallel one, which does not seem to me the case. ISTM that currently
Robins patches only update the parallel version.
--
Fabien.