> Otherwise LGTM, thanks.

OK, I think I have it almost ready see the patches below.
I just hit one snag when trying to fix the tests.

We have for example the following test:

    (ert-deftest pp-print-quote ()
      (should (string= (pp-to-string 'quote) "quote"))
      (should (string= (pp-to-string ''quote) "'quote"))
      (should (string= (pp-to-string '('a 'b)) "('a 'b)\n"))
      (should (string= (pp-to-string '(''quote 'quote)) "(''quote 'quote)\n"))

This is how the old code behaved, i.e. the output sometimes ends with \n
and sometimes not, depending on whether the object printed is a list or not.

Currently, my new code behaves the same when using `pp-28` or `pp-29`
but when using the new default (i.e. `pp-fill`) the output never ends in
\n.  This change was not intentional, but I think it makes sense because
it's more consistent.

I'm not completely sure how we should fix this.  I think the old
behavior of sometimes adding \n and sometimes not is not desirable, so
I think we should change it (a backward incompatible change).
We have two remaining choices:

A) never add \n
B) always add \n

AFAICT, in practice the old behavior resulted in a \n added in most
cases, so (B) should lead to less breakage, but OTOH I think (A) would
be cleaner since it's easier for callers to add a \n when needed than
for them to remove a \n.

WDYT?  A or B?


        Stefan