Re: OT: compile problem--solved (apparently)

[Kenward Vaughan <kaynjay@igalaxy.net>]

On Fri, Apr 27, 2001 at 11:44:55PM +0200, Erdmut Pfeifer wrote:
> On Thu, Apr 26, 2001 at 11:54:31PM -0700, Kenward Vaughan wrote:
> 
> > I hope someone can give me clues about what's screwing up the compilation of
> > a molecular docking app I'm trying to assemble for my classes... It is
> > supposed to compile on a number of *nix platforms including Linux, but my
> > attempts die at the linking stage.
> > (...)
> > 
> > Can someone suggest places to look for the cause of this?  Is there other
> > information I could find to help?  I have tried the nm command with the
> > three functions referenced below, and the only consistent difference I can
> > see is the lack of a bunch of alphanumerics after the name shown on the last
> > line of that listing.
> 
> this small difference in the symbolnames _could_ be an indication that
> there's some kind of prototype mismatch. You might want to use nm's -C
> option to demangle those cryptic alphanumerics. This will give you a more
> human-readable output, though I'm not sure whether it'll help you beyond
> that... (see 'man c++filt' for a short intro on what mangling is about).

The problem turned out to be in the get_atom_type.cc file.  The top lines
are:

/* get_atom_type.cc */


/* #ifdef sgi */
    #include <stdio.h>
    #include "get_atom_type.h"
/* #else
    extern "C"
    {
        #include <stdio.h>
        #include "get_atom_type.h"
    }
#endif  */

Note that I commented out all but the includes for an SGI.  (Don't ask me
why/how that works... I stumbled onto it.  I do thank Hendrik Sattler for
some prodding and another offer to help from Anthony Lau.)

 
> I would try to locate the implementation of the 'get_atom_type(char *, char *)'
> function in the source code (most probably in get_atom_type.cc) and see
> whether it has the required prototype (the one that the linker's "undefined
> reference" message complains about).


I guess this is what the above lines involve?



> Somehow, however, I doubt that some bug like this would have made it
> into the AutoDock distribution. Or is this a brand new release, that no
> one else has yet tried to compile?


I'm not sure.  There are several packages at the ftp site which appear to be
transitional ones.  My guess is this (the "current" one) still fits that
mold.

The group seems to focus largely on SGI, Alpha, Sun, HP, and Convex.  The
calculations are quite intensive... perhaps they don't have many requests
for Linux/PC machines?  There are some lines commented for Linux in the
Makefiles, but obviously there are other problems.

I'm admittedly a bit miffed by their complete lack of response in any way to
my emails.  Maybe I'm too sensitive??  :)

 
> Too bad this isn't open source -- otherwise I could have given it a
> quick try myself... Anyway, good luck!


Thanks.  I've already found another problem in a different program in the
package.  Different post for that one.
 
Kenward
-- 
It is not so very important for a person to learn facts.  For that he
doesn't really need a college education, for he can learn them from
books.  The value of an education in a liberal arts college is not the
learning of many facts but the training of the mind to thinking--something
that cannot be learned from books.     Albert Einstein