[CM] misbehavior in functions to shift and stretch regions

[bil at ccrma.Stanford.EDU]

The free cell problem probably means some s7 object
is not GC protected when it should be.  I haven't
managed to hit the bug with your code yet (GC bugs
depend on just about everything), but if you'd like
to poke at it, first build s7 with S7_DEBUGGING=1
(either through CFLAGS in a makefile, or define it
in s7.c), remake snd, run snd under gdb, hit the
bug, and the stacktrace should tell you which
pointer is unhappy; look at it via p ptr[0] in
gdb, and at the bottom of the copious output you'll
find where that cell was last allocated, and where
the gc call happened that freed it.  Sometimes that
is enough to fix the bug.

I haven't looked at the second bug yet, but at first
glance it looks like an optimizer problem.