[Coin-discuss] hopefully the COIN-OR build process doesn't reflect the quality of the codes themselves

[Brady Hunsaker]

The question of build tools came up recently on the list for COIN-OR 
maintainers.  My sense was that the same tradeoff was discussed: 
ease/familiarity for users versus complexity/unfamiliarity for 
developers.  No conclusive decision was made, but I suspect that some of 
the projects will be moving to autoconf/automake in the future.  If any 
of you have some expertise to offer it would be appreciated.

The other issue--that of official "stable" releases--has also been 
discussed since the recent formation of the COIN-OR Foundation.  I know 
that some of the members of the Technical Leadership Council are in 
favor of implementing this.  I can't put a timeline on it, but I believe 
this is something that will happen.  As Jean-Sebastien said, that will 
make creating ports/packages for distributions much more realistic.

Brady Hunsaker
Secretary, COIN-OR Foundation Strategic Leadership Board

Michael Hennebry wrote:
> On Sun, 20 Mar 2005, Jean-Sebastien Roy wrote:
> 
> 
>>I did not meant to say that the tools provided in COIN are not "stable"
>>themselves (even if they are evolving quickly), just that there is no
>>*non-changing* tarball of COIN available on coin-or.org which would
>>enable to make a FreeBSD port: I can't make a port that fetches a
>>different source tarball every day, nor a port that checkout a CVS
>>repository, nor I can host such a tarball myself.
> 
> 
> The bit about smoke was a joke.
> I understood not "stable" to mean "changes a lot".
> That said, sometimes things get checked in without being tested,
> sometimes without even being compiled.
> 


-- 
Brady Hunsaker
Assistant Professor
Industrial Engineering
University of Pittsburgh
http://www.engr.pitt.edu/hunsaker/