--- Begin Message ---
Subject: |
Emacs-28: Inadequate coding in hack-elisp-shorthands |
Date: |
Fri, 1 Oct 2021 17:10:57 +0000 |
Hello, Emacs.
In emacs -Q in the emacs-28 branch, create the following two line file,
foobar.el, and try to load it:
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
(defvar foo-baz "foobar-baz")
FOOBARELISP-SHORTHANDS: (("foo" . "foobar")))
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
This will throw an error, but that isn't important.
What is important is that the symbol foobar-baz is created by the
elisp-shorthands facility.
This shouldn't happen since:
1/- There is no Local Variables section.
2/- There is no variable elisp-shorthands in that non-existent section.
The following errors are evident in hack-elisp-shorthands:
1/- The code doesn't check for a correctly formatted Local Variables
section.
2/- The code, even if it did check, would only check the last 3000 bytes
in the file. The section can occur anywhere in the last 3000
CHARACTERS.
3/- The code doesn't do a case-sensitive search for "elisp-shorthands".
4/- The code doesn't check for "elisp-shorthands" being a complete
symbol.
5/- The code doesn't even check that "elisp-shorthands" is in a comment.
I would suggest that these errors be corrected. I would also suggest
that the entire code and documentation for this new facility be
carefully reviewed by somebody who isn't the original author.
--
Alan Mackenzie (Nuremberg, Germany).
--- End Message ---
--- Begin Message ---
Subject: |
Re: bug#50946: Emacs-28: Inadequate coding in hack-elisp-shorthands |
Date: |
Sat, 02 Oct 2021 12:13:14 +0100 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/28.0.60 (gnu/linux) |
Alan Mackenzie <acm@muc.de> writes:
> Why aren't you just using that function on the buffer anyway, instead of all
> this
> clumsy messing around with temporary buffers, file-attributes, and
> successive 100 bytes, and so on?
I used it to quickly and efficiently discover the value of
elisp-shorthands as soon as possible, before the 'read' takes place in
load-code-with-conversion.
Doing other wise is possible but it is an intrusive change into the
longstanding load-with-code-conversion, which I avoided during the early
days of this feature. If you can do it and it doens't break anything,
go ahead. Stefan is investigating this as well.
> ... even if this scenario is highly unusual.
You think?
João
--- End Message ---