[Coin-discuss] BCP and Windows

Brady Hunsaker hunsaker at engr.pitt.edu
Fri Oct 31 08:57:58 EST 2003


On Fri, 2003-10-31 at 08:39, Stephan Hennig wrote:
> Sorry Brady, this should have gone out to the list. But fortunately,
> know that I'm less frustrated I changed the posting slightly.
> 
> Brady Hunsaker schrieb:
> > 
> > A quick search of the coin-discuss archives shows no occurrences of
> > the word "Borland" and I don't remember anyone ever mentioning it.
> > So probably no one has reported on efforts to try Borland with BCP.
> 
> Which makes me very sad. Borland C++ Builder is such a powerfull and
> yet user-friendly IDE, it's fun working with it.
> 
> 
> > We'd certainly appreciate if you let us know how your efforts go.  I
> > guess issues that have come up before are related mainly to
> > templates, though I could be wrong.
> 
> Unfortunately, it doesn't compile coin. And I'm not a system programmer
> who can apply the necessary adjustments. That's probably beyond my
> skills. I just wanted to use the API.
> 
> As the very first problem BCB can't find <unistd.h> on line 11 in file
> CoinHelperFunctions.hpp. There is no such file on my machine. I can hear
> you laughing, but I have no idea how simple/complicated that problem is.
> 
> Is there any hint to what problems would/could/should raise up and have
> to be dealt with. What were the critical files when doing adjustments to
> make COIN run with MSVC++ (which is no option for me)?
> 
> Has really no-one tried BCB before (with a little more system
> knowledge)? If not, I have to turn to linux (at the moment it seems
> Symphony makes troubles on Windows too) and hope to find some suitable IDE.
> 
> Kind regards,
> Stephan Hennig
> 

Stephan,

I was just composing a reply to your email when I received this
replacement.

I'm pretty sure no one has communicated with the list about efforts to
compile COIN-OR software using Borland.  There have been successful
efforts to compile COIN-OR code under MS Visual C++, however.  On the
assumption (??) that Borland and MSVC++ are somewhat close, you could
try adding the line
#define _MSC_VER
before even the include files.

This is used in some places in the COIN-OR code to handle compatibility
issues, including the <unistd.h> problem.  Whether or not this helps
depends on whether MSVC++ and Borland code is easily interchangeable.  I
have no idea whether or not that is true.  

Unfortunately, BCP is complicated enough that I will be (pleasantly)
surprised if this is sufficient.  You can search for "_MSC_VER" in the
code to find the places where there have been issues in the past, if you
are willing to try to adjust them for Borland.  It's possible there will
be new issues as well.

I do hope that we will make more efforts to support Windows in the
future, but even then, I expect that we will target support for
open-source compilers, such as gcc under cygwin, or MinGW.

Can anyone right now indicate the level of support for MinGW or
gcc/cygwin?

Brady




More information about the Coin-discuss mailing list