Re: crash in 1.0.3a fortran compilation

[Craig Burley <burley at gnu dot org>]

>Probably worth doing.  I'm not sure we'd want to include this second
>tier of tests in the distributions though.

I think it's probably best to *not* include them, but include some
kind of (casual?) data base on what they are (name and version #s),
where they can be obtained, what is known in terms of procedures to
test them (the better to automate this), etc.

That is, we can start collecting this info *now*, and try to put it
into a form that is not only readable by humans, but might be
automatable, so someday a `make full-check' target might exist that'd
see if it had to download the package first (if it wasn't on the
system already), then run the test procedure, etc.

In other words, I prefer the approach of collecting and providing a
little bit of information (URLs and procedures) over a whole ton
of data (entire distributions that are already available from other
sources).

        tq vm, (burley)