[ghemical-devel] ghemical/libghemical works again, +some improvements

[Tommi Hassinen]

On Fri, 24 Jan 2003, Michael Banck wrote:

> Well, the first problem I ran into was that I didn't have f2c.h
> installed. The configure-script should probably exit with an error if it
> cannot find it.

Yes, I have also seen some cases where the configure-script should in fact
stop when something critical is missing.

But please remember that I'm not very good at handling these configuration
issues; my expertise is limited to running "autoconf" and "./configure"
and making minor changes to the related files. So if you know any
improvements to the configuration system, I'll be happy to include them.

> But this leads to another question: If you factor out libghemical,
> wouldn't it make sense to give MOPAC a seperate source tree, too?

The only problem, or question, here is that the original MOPAC7 source and
the built-in "miniMOPAC" are a bit different. Since the "miniMOPAC" is
linked into the ghemical executable, it's original MAIN function is left
out and replaced with an different function. The files are:

	mopac.f / mopac.c		the original
	minimopac.f / minimopac.c	modified

So, a stand-alone MOPAC (and a library?) should contain the original
files, and in ghemical we only should replace the MAIN function with our
own version.

Probably this can be done, but like I'm not good at handling configuration
files, I recently found out that I'm not good at libraries either.  :)
I tried to use separate libghemical.a and libminiMOPAC.a (within the
current source tree) but even that turned out to be too complicated.

> In fact, I've met one person at our uni who had quite a hard time
> messing with it and trying to get it compiled on Linux (I pointed him to
> ghemicals version of MOPAC, I don't know it that helped or not). I don't
> know if there would be many people around interested in maintaining the
> source (I know that I don't have the time and expertise in Fortran77),
> but at least this would clean up libghemical considerably.

The result of my experiments is that the f2c tool is the best and safest
way to run MOPAC.

> I started autoconfiscating MOPAC7 a year ago or so, but I can't find
> this anymore on my harddisk at the moment. Anyway, it doesn't look
> overly complex, so I could do it again if you decide to split MOPAC off.

The above problem with the mopac MAIN function is the only one I know
that would be related to the split. Any ideas about it?

Regards,

	Tommi