🚀 go-pugleaf

Larry Brasfield wrote:


> Your query would be better directed to gnu.g++,
> where you would find a higher proportion of
> folks familiar with that compiler and linker.

Thanks for the tip.  I'll try that.

> >libtmp/_spring.o: In function `acosh':
> >libtmp/_spring.o(.text+0x418): multiple definition of `acosh'
> >libtmp/_ABMP_matching.o(.text+0x418): first defined here
> >
> >Mostly complaining about symbols from the math library (math.h)
> >that are multiply defined.
> 
> You are linking inappropriate libraries.  I
> don't know why you're linking that stuff,
> but you should spend some time with the
> "nm" utility, learning what is to be found
> defined in each library.  Hint: It is easier
> to sort out a few missing definitions than
> to cure scads of duplicate ones.  It helps
> to have got the definitions listed via nm
> into a set of text files that you can grep.

First, some background.  LEDA is a C++ library of data structures
and algorithms.  It's well used, documented, and tested, and
comes with a configuration utility and makefiles for a number of
different platforms/compilers including Linux/g++.  The errors
that I posted here occur when I compile the demo code that comes
with the libraries, after, of course, compiling the libraries
themselves.  The files spring.o and ABMP_matching.o are object
files included in the libraries.

The function acosh that the linker is complaining about in the
above output is part of the standard C math library, linked in by
passing the -lm argument to the linker.  I've looked in the files
spring.c and ABMP_matching.c and acosh is definitely not
redefined in these files, but math.h is #included in these
files.  So, for some reason, it looks like just #including'ing
math.h is causing these symbols to be defined in the object
files.

Normally, I would think that this is a problem with the
Makefiles, but LEDA is pretty well tested, and a problem this big
would have been fixed long ago.  The compiler and linker I'm
using is the stock g++ that comes with Red Hat Linux 5.2, so that
shouldn't be the problem either.  Since I don't think that my
situtation is unique, I was hoping that someone else might have
run into exactly the same problem and been able to give me a one
line solution. No luck so far.

Thanks anyway,

Pat