bug#50946: Emacs-28: Inadequate coding in hack-elisp-shorthands

[Alan Mackenzie <acm@muc.de>]

Hello, Eli.

On Sat, Oct 02, 2021 at 18:00:38 +0300, Eli Zaretskii wrote:
> > Date: Sat, 2 Oct 2021 14:45:52 +0000
> > Cc: joaotavora@gmail.com, 50946@debbugs.gnu.org
> > From: Alan Mackenzie <acm@muc.de>

> > > > The five aspects I enumerated on my original bug report.  Not checking
> > > > for a properly formatted Local Variables: section

> > > That is not part of the function in question, is it?  It's in
> > > hack-local-variables--find-variables, which we use everywhere.

> > It is now.  It wasn't when I raised the bug a day or two ago.

> So this issue is no longer pertinent, right?

It was pertinent to my observation, but is no longer so, given that you
have reviewed the new C code.

[ .... ]

> > > > not going back at least 3000 characters

> > > That is now fixed, right?

> > No, it's not.  In certain edge cases, it will go back fewer than 3000
> > characters.

> Does the patch below solve this?

I think it does, yes.  Thanks!

> > > > I worry, to a lesser degree, it is not entirely clear whether setting
> > > > the elisp-shorthands variable in the first line of a short file should
> > > > be valid or not.  I don't think the current hack-elisp-shorthands is
> > > > careful enough about this.

> > > Why does it matter?

Otherwise we could have the scenario where somebody sets elisp-shorthands
in the first line of a file, finds it works, then types more into the
file, saves the buffer, then finds when she visits the file again that it
no longer works.  This, I think, would be a Bad Thing.

> > Because the first line definition should either be valid or not valid.
> > Currently it works for a sufficiently small file, but not for a normal
> > sized file.  This, I think, is a bug.

> No, I don't think it's a bug, at least not a bug specific to
> shorthands.  That's how file-local variables work in general.

No, not quite.  For normal file-local variables, having one set in the
first line works regardless of the length of the file.  It wouldn't for
elisp-shorthands, where it would only work for short files.

> > Say you have a file 3150 bytes long, which is less than 3000 characters
> > in Emacs.  Your function will load only 3100 bytes, less than 3000
> > characters, into the temporary buffer.  It thus may fail to find a Local
> > Variables section, even if this scenario is highly unusual.

> This should be solved by the change below.

Yes, thanks.

> > Have you checked that things work if the first byte in your temporary
> > buffer isn't at the start of a character?

> I don't see why this matters, can you explain?

It might matter, I simply don't know.  I rarely type characters into
Emacs which are longer than a single byte in UTF8.  I don't know whether
insert-file-contents does the Right Thing when there's half a character
at point-min, then insert-file-contents inserts the other half of the
character before it.  I don't know to what extent normal Emacs functions
work when there are invalid "characters" at point-min or point-max.  If I
were writing this function, I would want to check these things.

You are an expert on Unicode, so you are far likelier to know how Emacs
handles such things.

> Here's the patch I promised:

> diff --git a/lisp/shorthands.el b/lisp/shorthands.el
> index b8204d6..6162efd 100644
> --- a/lisp/shorthands.el
> +++ b/lisp/shorthands.el
> @@ -40,7 +40,10 @@ hack-elisp-shorthands
>      (with-temp-buffer
>        (while (and (< (buffer-size) 3000) (>= from 0))
>          (insert-file-contents fullname nil from to)
> -        (setq to from from (- from 100)))
> +        (setq to from
> +              from (cond
> +                    ((= from 0) -1)
> +                    (t (max 0 (- from 100))))))
>        ;; FIXME: relies on the `hack-local-variables--find-variables'
>        ;; detail of files.el.  That function should be exported,
>        ;; possibly be refactored into two parts, since we're only

Thanks!  I think it's right.

-- 
Alan Mackenzie (Nuremberg, Germany).