[Coin-discuss] OsiDylp --- what's it been like?

[Matthew Saltzman]

On Thu, 2 May 2002, Lou Hafer wrote:

> 	Pretty much none of this has shown up in COIN-discuss. About the only
> hint you have seen was a message a few weeks ago asking about experience with
> GCC compiler versions. In the main, I kept this off the list because it would
> be mostly unintelligble (and hence uninteresting) to those not actively
> working on dylp or the osi and common components of COIN.

An update on that message (referring to failure of the OSI unit test with
the Red Hat gcc 2.96 compiler from Red Hat Linux 7.2):

I was working on creating a "short" code that illustrates this problem,
but in the meantime I upgraded to the new Red Hat Linux 7.3
(gcc-2.96-110). The OSI unit test now runs correctly (at least past the
point of the earlier crash).

The bug was associated with the exception handler, and it turned up in a
place where correct code would not throw an exception.  It was also highly
location-dependent: if certain modules were removed from the unit test,
the problem disappeared (so I was able to create a short execution path,
but a lot of code needed to be included to get things in the right
location for the bug to show).

So is the bug fixed?  I don't know.  I didn't get to file it and now I
can't reproduce it.  Is the compiler acceptable for use with COIN-OR?  I
think so.  It is probably at least as good as 2.95.3 (CPLEX's opinion to
the contrary notwithstanding).  There's no way of knowing if the bug is
absent in the older compiler, but it doesn't seem that likely.  And the
unit test that failed before now does not.  So Red Hat Linux users should
upgrade.

-- 
		Matthew Saltzman

Clemson University Math Sciences
mjs at clemson.edu
http://www.math.clemson.edu/~mjs