ngspice

inpkmod.c modified by initializing modtab:

/* global input model table. */
INPmodel *modtab = NULL;

See comment to bug no. 2874418

Holger

Charles,

the second one is already in CVS.

I will test your patch and upload it to CVS.

Thanks

Holger

Aha, looks like the first patch here fixes things by initializing the
pointer to NULL:

http://trac.macports.org/browser/trunk/dports/science/ng-spice/files

Not sure if the second one has been merged into CVS yet, but I included it
in the Fink package just in case.

Sorry, I found the definition in ./src/spicelib/parser/inpmkmod.c.

I still wonder if GNU ld is doing something different, like sorting
things.

Also, I am looking through the CVS source, and I don't see how modtab is
being defined (except as an extern pointer). I wonder if the code is
relying on a feature of GNU ld?

If the ngspice-19 tarball was generated from CVS, then running 'cvs rdiff'
between the proper -19 tag and HEAD should yield a patch file that, when
applied to the tarball, is equivalent to checking out HEAD from CVS. (I am
trying to minimize the potential differences between building by hand, and
building the way the Fink package builds.)

Some new problems have cropped up (related to readline?) so I will need to
go chase those down first.

Charles,

the best would be if yout check out the code from CVS and test it.

Thats what I am working on. To follow the "old" tarball in detail would be
difficult and perhaps does not lead to a solution.

We might consider making a new release.

Regards

Holger

Holger,

In my build process, I am using the released ngspice-19 tarball plus a
patch file generated from CVS. Which tag should I diff from, rework-19 or
rework-19_2?

Charles,

the code work under mingw, cygwin, LINUX (Suse11.2).

Still I have made a small change and uploaded to CVS. Please try again.

Regards

Holger