[PATCH] unsafep: Add support for (funcall)

[PATCH] unsafep: Add support for (funcall)

[Vibhav Pant]

The following patch makes (unsafep) check funcall forms. Right now,
it only works while calling side-effect-free functions like car and cdr.
In the future, I'd like to extend this to add support for looking up and
checking (symbol-function) for other functions too.

Would appreciate some feedback before I push this.


Thanks,
Vibhav

-- 
Vibhav Pant
vibhavp@gmail.com

diff --git a/lisp/emacs-lisp/unsafep.el b/lisp/emacs-lisp/unsafep.el
index 1ab65a044e..beba6895b8 100644
--- a/lisp/emacs-lisp/unsafep.el
+++ b/lisp/emacs-lisp/unsafep.el
@@ -205,6 +205,12 @@ unsafep
        ((eq fun '\`)
  ;; Backquoted form - safe if its expansion is.
  (unsafep (cdr (backquote-process (cadr form)))))
+       ((eq fun 'funcall)
+        (let ((func (nth 1 form))
+              (args (nthcdr 2 form)))
+          (if (eq (car-safe func) 'quote)
+              (unsafep (cons (eval func) args) unsafep-vars)
+            (setq reason `(function ,func)))))
        (t
  ;;First unsafep-function call above wasn't nil, no special case applies
  reason)))))

Re: [PATCH] unsafep: Add support for (funcall)

[Stefan Monnier]

> The following patch makes (unsafep) check funcall forms. Right now,
> it only works while calling side-effect-free functions like car and cdr.
> In the future, I'd like to extend this to add support for looking up and
> checking (symbol-function) for other functions too.

Hmm... I don't quite get it:

    (funcall '<sym> ...)

is better written as

    (<sym> ...)

so why/when do you see such code?  Could we rewrite it before unsafep
gets to look at it.


        Stefan

Re: [PATCH] unsafep: Add support for (funcall)

[Vibhav Pant]

On Tue, Feb 28, 2017 at 2:26 AM, Stefan Monnier
<monnier@iro.umontreal.ca> wrote:
> Hmm... I don't quite get it:
>
>     (funcall '<sym> ...)
>
> is better written as
>
>     (<sym> ...)
>
> so why/when do you see such code?  Could we rewrite it before unsafep
> gets to look at it.

Indeed, byte-compile-funcall rewrites such forms. I've just added this to remove
a potential false positive from unsafep, as it's not only the byte compiler and
optimizer that use it (for instance, ses.el uses it to warn the user
about "unsafe"
formulae).


-- 
Vibhav Pant
vibhavp@gmail.com

Re: [PATCH] unsafep: Add support for (funcall)

[Stefan Monnier]

>> Hmm... I don't quite get it:
>> 
>> (funcall '<sym> ...)
>> 
>> is better written as
>> 
>> (<sym> ...)
>> 
>> so why/when do you see such code?  Could we rewrite it before unsafep
>> gets to look at it.

> Indeed, byte-compile-funcall rewrites such forms. I've just added this
> to remove a potential false positive from unsafep, as it's not only
> the byte compiler and optimizer that use it (for instance, ses.el uses
> it to warn the user about "unsafe" formulae).

But unsafep must call something like macroexpand-all or something like
that, so we could (should?) arrange for that to do the rewrite.


        Stefan