[Coin-discuss] Status of COIN interface to bonsai

[Laszlo Ladanyi]

Lou,

I'm willing to track down what's needed for g++-3.0. The only problem is that
g++ support in gdb is a bit flakey even for the g++-2.95, and gdb has not yet
been updated to handle the slightly different mangling of g++-3.0... 
So the development environment should be g++-2.95 for now, but you are right,
we should make it so that it compiles with g++-3.0. After informs I'll attack
the problem. Of course, everyone is welcome to fix it earlier :-).

--Laci

On Fri, 26 Oct 2001, Lou Hafer wrote:

> Brady, & others,
> 
> 	What's actually in the works is a COIN interface to dylp, the lp
> solver embedded in the bonsai MILP code. Stephen Tse did the development over
> the summer, then started working on a different project as of September.
> There was a hiatus through September and most of October while Stephen and I
> concentrated on other things. The current status is that Stephen and I are
> wrestling with two development environments and three compilers, trying to
> find some compatible version of code. (And, though I'm a bit abashed to admit
> it, progress is slowed because this is my first real project with C++ :-).
> 
> 	I work in a Solaris/Sun Workshop programming environment. Sun farmed
> out Workshop development to a company called Forte; currently we're at
> edition 6 update 1.  Stephen works in a Solaris/Gnu environment.  Our
> computer systems folks recently upgraded us from Gnu 2.95 to Gnu 3.0.
> 
> 	The COIN interface for bonsai was originally developed by Stephen
> using Gnu 2.95 tools. We can also build COIN itself using Gnu 2.95. So we
> have working code for the Gnu 2.95 tool set. And probably should have left it
> at that.
> 
> 	But ... right now we can't build COIN in the Workshop environment,
> nor can we build it using Gnu 3.0. For obvious reasons, I'd rather be able to
> work with COIN in the Workshop environment, and it strikes me that this would
> be a nice goal on general principles. The basic problem seems to be that the
> Forte C++ compiler is less bright than g++ when it comes to matching actual
> code to specialisations of templates. I've been able to beat the problems in
> our code, but am just now getting down to work on the COIN code.
> 
> 	The problems with Gnu 3.0 are, I think, transitory --- the autodetect
> seems to be doing something silly. The appearance is that it's lost a basic
> library file, leaving reams of undefined symbols. It's a matter of spending
> some time and looking carefully at how the makefiles actually work.
> 
> 	If anyone has successfully built COIN in Sun Workshop, or experienced
> similar problems with Gnu 3.0 under Solaris, we'd be interested to hear.
> 
> 						Lou 
> _______________________________________________
> Coin-discuss mailing list
> Coin-discuss at www-124.ibm.com
> http://www-124.ibm.com/developerworks/oss/mailman/listinfo/coin-discuss
>