Philipp Stephani <p.stephani2@gmail.com> writes:
> Thanks for this great analysis. Given this, it seems that the place
> definition for `buffer-local-value' should be removed from gv.el.
But hmm - surely there are other functions where `cl-letf' doesn't
exactly restore the previous state - `alist-get', for example:
#+begin_src emacs-lisp
(setq my-alist '((x . 1)))
(ignore (cl-letf (((alist-get 'y my-alist) 17)) my-alist))
my-alist
==> ((y) (x . 1))
#+end_src
(admittedly, this is not as serious as the `buffer-local-value' case).
Also, as another, different problematic case, the manual warns about
`point' to be used with `cl-letf'.
So, I wonder if, instead of removing the gv-setter definition for
`buffer-local-value', we should instead add some more text about how
`cl-letf' can have surprising effects to the manual.
I think we should spend significant efforts to avoid surprises. In this case, if it means we should remove `alist-get' as well from the forms supported by `cl-letf', then I think that's what we should do. The documentation for `cl-letf' clearly states: "On exit, either normally or
because of a ‘throw’ or error, the PLACEs are set back to their original
values." If it can't do that for some place form, it shouldn't be allowed.