Re: [PATCH] Fix hack-local-variables for find-file-literally with dos encoding

[Lars Ingebrigtsen]

Tom Gillespie <tgbugs@gmail.com> writes:

> No, when we read a file literally we don't want to normalize anything.
> The issue is that 'when noconv' is set (i.e. we are in find-file-literally)
> then we don't want to convert ?\^m because the value of suffix
> contains ?\^m. Does this make sense? I tried other approaches
> including modifying the suffix but this one was by far the cleanest.

Ah, I see what you mean now.  The code in
`hack-local-variables--find-variables' is clear enough in what it does,
but it's rather obscure in what it intends to do.

The issue is that if you've loaded a DOS file literally, and then you
want to run the local variables anyway, then you have a file like this:

;; -*- mode: Emacs-Lisp -*-^M
;; Local Variables:^M
;; lol: t^M
;; End:^M

Then we have this:

      (when (let ((case-fold-search t))
              (search-forward "Local Variables:" nil t))
        (skip-chars-forward " \t")
        ;; suffix is what comes after "local variables:" in its line.
        ;; prefix is what comes before "local variables:" in its line.
        (let ((suffix
               (concat
                (regexp-quote (buffer-substring (point)
                                                (line-end-position)))
                "$"))

So...  it notes all characters that are after "local variables:", which
in this case is ^M.

Then later it does

            (with-temp-buffer
              (insert-buffer-substring thisbuf startpos endpos)
              (goto-char (point-min))
              (subst-char-in-region (point) (point-max) ?\^m ?\n)
              (while (not (eobp))
                ;; Discard the prefix.
                (if (looking-at prefix)
                    (delete-region (point) (match-end 0))
                  (error "Local variables entry is missing the prefix"))

and this fails.

I think the intention here is to be able to say:

;; Local Variables:my-identifier
;; lol: t
;; End:my-identifier

although that's not something I can remember seeing in the wild.

So I think your revised patch is correct -- the `subst-char-in-region'
thing is just plain wrong and should be removed.

Does anybody else have any insights into this corner of the
hack-local-variables code?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no