I think I found what's wrong. It is not the package as a whole.

I used the exapmles to check whether the rpm builds were OK. I thought
that if the examples compiled OK then the package would be OK. I
suspected that it might be a SMP problem, and it might be. But not when
it comes to compile the package, only when it comes to compile the
examples. I think that the pagkage was built OK every time. It was just
the examples that were failing

I think it is last few lines in Makefile.am that is failing

foo-glue.c: foo.defs
        build-guile-gtk glue $< >tmp && mv tmp $@

main.c: foo.defs
        build-guile-gtk main $< >tmp && mv tmp $@

With this above in Makefile.am a build would fail 9 times of 10 or so.

I dont really understand how this works but I think the tmp-something is
deleted before the final file is ready written and that it is here it
might be an SMP problem. I am building on a dual CPU box.



Whith this  in Automake.am it works. I have tried several builds.

foo-glue.c: foo.defs
        build-guile-gtk glue $< >$@

main.c: foo.defs
        build-guile-gtk main $< >$@


Here is the last src.rpm
ftp://geek.mine.nu/pub/SRPMS/guile-gtk-20011121-8.src.rpm

The latest patch comes as an attach

G'nite
Bo