urgent bug report/help requested

[Thomas Bushnell BSG]

I'm pleased to report that I located the problem.  The assistance of
all was appreciated; even the negative cases where people *didn't* see
problems helped me rapidly narrow down the things to check.  The bug
is a bug in slib, which is easily patched and I'll submit the change
to Aubrey and fix Debian today.

What follows is a more detailed description, for those interested, and
in case it recurs for anyone:

slib's portability across schemes is guaranteed by a series of .init
files, one for each scheme system out there.  slib comes with these
.init files.  For a long time however, guile's slib interface has not
used the slib-provided guile.init file, and instead the guile
developers took the then-current guile.init file, hacked it, and
installed it as slib.scm.

Then, each time slib was released, they would play catchup (or not),
and over time, slib.scm and guile.init drifted further and further
apart.  With the incorporation of slib-3a2 into Debian, making this
work was more than it was worth, and we switched to having guile load
directly the slib guile.init file.  Guile upstream also agreed that
this was the right long-term course.

Aubrey and the guile people tried to make sure that guile.init
correctly did the right thing, buc since guile.init had been unused
for years (essentially unknown to Aubrey!) they missed one.  There is
crucial module-flummery that must be done inside the slib:load
function, to make sure that when slib loads something, it lands inside
the slib module, and not just whatever module happens to have invoked
require.  

The fix is simple, and I'm making it in Debian today and mailing it
off to Aubrey as well.

Thomas