cw has never been NULL here,
because we used to survive an update of cw->wl_next here
wlist hasn't been NULL either,
because there is no place where wlist becomes NULL without
setting cw to NULL as well. yet cw isn't NULL (see above)
aproach the `append'
this modification is faithful iff cw != NULL and cw->wl_next == NULL
cw != NULL because we have just survived an update of cw->wl_word
cw->wl_next == NULL is witnessed by:
1) cw either NULL or pointing to a wl_cons(,NULL)
2) cw not being modified, and no wl_next being modified
3) check of all function invocations with wlist/cw argument
not to modify any wl_next slot
which are:
1) cp_ccom(wlist, ...)
2) wl_free() always followed by goto nloop, which does a cw = wl_cons(,NULL)
and cp_ccom() doesn't modify any wl_next slot
the first part of the patch is intentionally similar to newword
`i' and `buf' can be modified here, because we will inevitably `goto done'
attach and instantly remove a NULL wl_word here
rlar
dwarning
h_vogt
pnenzi
arno