[Cbc] Proposed change to cbc unit test

[Ted Ralphs]

The reason we created Data/Sample and made the unit test depend only
on instances there was to make the download footprint smaller and to
maintain the ability to run just a quick "installation verification
test," not really intended to be a full unit test. The long-term plan
was to have more complete tests that one could execute by, for
example, doing "make fulltest". This is the way things are implemented
in SYMPHONY at the moment. In fact, what we do there is that if the
user types "make fulltest", then additional miplib instances are
automatically downloaded at that time if they are not already present.
This way, users that just want the IVT don't have to have the extra
instances around if they don't want. With disk sizes and bandwidth
increasing, perhaps the original arguments against making the full
miplib a requirement to run the unit test don't apply anymore.
However, I would still argue that we should stick to our original
philosophy of having "make test" just run a basic IVT and implementing
a more complete test separately. If you think flugpl would be a better
instance for an IVT, then I would vote to just add it to Data/Sample.

Cheers,

Ted

On Fri, Mar 25, 2011 at 12:44 PM, Lou Hafer <lou at cs.sfu.ca> wrote:
> Folks,
>
>        The cbc unit test currently solves p0033 and p0201, both pure binary
> IPs. I'd like to add flugpl, which is small but has general integer and
> continuous variables. This provides better test coverage. Flugpl is not
> part of Data/Sample. The easy solution for cbc is to point the unit test
> at Data/Miplib instead of Dir/Sample. The alternative is to add flugpl
> to Data/Sample.
>
>        I will point the unit test at Data/Miplib if no one objects within a
> day or two. If anyone has other suggestions for unit test problems (they
> should be small and solve quickly), speak up.
>
>                                                Lou
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
>



-- 
Dr. Ted Ralphs
Associate Professor, Lehigh University
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted