This is far from complete, but I've had so little time to work on it
this week, and wily's had so many releases, that I thought I may as
well give others a chance to play with it. Besides, I was getting fed up
with applying that patch to text.c.

This includes regexp.c, but again, it's unchanged from the original.
The sam.[ch] here are different from those in wily itself.

This works on my machine with the pipeline:

e file | x /regexp/ | c word | q

but I haven't played with many combinations yet. As long as the
text.c patch goes in, then it will be easier to test it, rather than
trying to keep up with wily revisions. Couple of major points:

- e asks wily for "." by default, if no address is given. During the
resolution of this, wily attempts to open "/path/." where e has
asked for "." in /path/name. This is kind of weird, and is part of
the context rules. I'm not sure what to do about it.

- "q" isn't really necessary, if we take acme's stance, and say that
you can only have one of c,i,a,d in the pipeline. Seems a little mean
to me, since they don't supply s. I've tried to arrange things so that
you can pipe c,i,d,a into each other, but that's not working yet. Or
we could just give up (combinations of i and a are ok, c and d don't
make a lot of sense in combination), add s, and dump q. Which will
probably for the best anyway.

- I haven't done anything about arranging things so that Wily allows
Undoing, yet. Haven't even looked into it.

Steve
