[PPL-devel] Compiling PPL under Cygwin?

[Roberto Bagnara]

Goran Frehse wrote:
 > Roberto Bagnara wrote:
 >
 >> I assume the library that could not be found is `libstdc++.la',
 >> not `libstc++.la'.  If so, this is the standard C++ library and
 >> should have been installed along with g++ on your installation.
 >
 > Yes, you are right.
 > Actually I just noticed that somebody answered my request in the Cygwin
 > mailing list about this. It turns out that the paths in the .la files
 > are wrong and must be corrected by hand in the current installation.
 > For details, see:
 > http://cygwin.com/ml/cygwin/2004-02/msg00304.html

Dear Goran,

thanks for the pointer.  I have tried myself and I can confirm that
the problem is the one indicated.  I hope they fix the problem soon,
since, with the current Cygwin, no version of the PPL can be built
without fiddling with the .la files.

 >> Meanwhile, a few questions: (1) which version of
 >> the PPL are you trying to compile?  (2) Which version of Cygwin
 >> are you using (if you use the updater http://sources.redhat.com/cygwin/
 >> then we are using the same version)?  (3) Did you compile GMP yourself
 >> and, if so, how did you configure it?
 >
 > (1) 0.5
 > (2) 1.5.7-1, a fresh download from www.cygwin.com last week.
 > (3) GMP was compiled without further options (It installed itself in
 > /usr/local/lib).

You need to configure GMP with --enable-cxx in order to enable its
C++ interface.

 > Note that "make" causes no problems, only "make check".

I know, the same happens under Linux.  I guess this is a GMP bug.

 >> I attach the current version of these instructions for the C interface
 >> under the signature.  I will draft something for the C++ interface
 >> and let you have it, so that you can suggest improvements and additions.
 >
 > Great. Please know that I'm not trying to cause you work, just trying to
 > be of assistance.

And you are being of assistance.  Moreover, causing work is good,
since doing things upon user's demand is a good way to schedule our
activities.  It is so because (1) the user can then check the result
with a fresh eye, and (2) several things we would like to do will
remain in the waiting list anyway.

 > It took me a few attempts to get everything to compile
 > under Linux, so a few small comments might help others, too.

Of course.  When you have a minute, please do send that comments to
us: we will include them in the documentation and/or web site.
Thanks a lot,

     Roberto

-- 
Prof. Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara at cs.unipr.it