bug#11258: 24.1.50; lexical binding inconsistencies

bug#11258: 24.1.50; lexical binding inconsistencies

[Michael Heerdegen]

Hi,

I have an elisp file with this content:

; Test  -*- lexical-binding: t -*-

(defun test1 (arg)
  (interactive "i")
  (funcall (lambda () arg)))

(defun test2 (arg)
  (interactive "i")
  ((lambda () arg)))

;; (test1 nil)
;; (test2 nil)

I visit the file and do M-x eval-buffer.  Then I do the following:

1(a) Put point after (test1 nil) and hit C-x C-e: nil
1(b) M-x test1: nil
2(a) Put point after (test2 nil) and hit C-x C-e
     If I had loaded my .emacs before, I get nil
     With emacs -Q, I get: Lisp error: (void-variable arg)
2(b) M-x test2: Lisp error: (void-variable arg)

Now, byte compile the file, and load the compiled code.  Now, repeat
the above four tests.  You always get nil, and no error.

That's strange - shouldn't the result always be nil?  If not, this
(surprising) behavior should be documented in the manual.

I've experimented a lot, and found that the results for 2(a) are not
always the same.

OTOH, the compiled code always works as expected (result nil).


Regards,

Michael.


In GNU Emacs 24.1.50.1 (i486-pc-linux-gnu, GTK+ Version 3.2.3)
 of 2012-04-10 on zelenka, modified by Debian
 (emacs-snapshot package, version 2:20120410-1)
Windowing system distributor `The X.Org Foundation', version 11.0.11104000
Configured using:
 `configure '--build' 'i486-linux-gnu' '--host' 'i486-linux-gnu'
 '--prefix=/usr' '--sharedstatedir=/var/lib' '--libexecdir=/usr/lib'
 '--localstatedir=/var' '--infodir=/usr/share/info'
 '--mandir=/usr/share/man' '--with-pop=yes'
 '--enable-locallisppath=/etc/emacs-snapshot:/etc/emacs:/usr/local/share/emacs/24.1.50/site-lisp:/usr/local/share/emacs/site-lisp:/usr/share/emacs/24.1.50/site-lisp:/usr/share/emacs/site-lisp'
 '--without-compress-info' '--with-crt-dir=/usr/lib/i386-linux-gnu/'
 '--with-x=yes' '--with-x-toolkit=gtk3' '--with-imagemagick=yes'
 'build_alias=i486-linux-gnu' 'host_alias=i486-linux-gnu'
 'CFLAGS=-DDEBIAN -DSITELOAD_PURESIZE_EXTRA=5000 -g -O2''

bug#11258: 24.1.50; lexical binding inconsistencies

[Stefan Monnier]

> ; Test  -*- lexical-binding: t -*-
[...]
> (defun test2 (arg)
>   (interactive "i")
>   ((lambda () arg)))

Yes, this is the inconsistency recently discovered in
emacs-lock-mode, indeed.

> That's strange - shouldn't the result always be nil?  If not, this
> (surprising) behavior should be documented in the manual.

This is a bug in the handling of ((lambda ...) args) in lexical-binding:
the compiler treats it as (funcall #'(lambda () ...) args), whereas the
interpreter treats it as (funcall '(lambda () ...) args), so the
lambda's body is treated as lexically scoped when compiled but
dynamically scoped when interpreted.

I think the compiler is right, so the interpreter needs fixing.  I'm not
completely sure it will be harmless, so it won't be fixed in 24.1.


        Stefan

bug#11258: 24.1.50; lexical binding inconsistencies

[Michael Heerdegen]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

> > ; Test  -*- lexical-binding: t -*-
> > (defun test2 (arg)
> >   (interactive "i")
> >   ((lambda () arg)))

> This is a bug in the handling of ((lambda ...) args) in lexical-binding:
> the compiler treats it as (funcall #'(lambda () ...) args), whereas the
> interpreter treats it as (funcall '(lambda () ...) args), so the
> lambda's body is treated as lexically scoped when compiled but
> dynamically scoped when interpreted.

There is no scoping at all when interpreted.  The dynamical scope is not
established (else, we would get no error), which is right - just the
lexical scope is missing.

> I think the compiler is right, so the interpreter needs fixing.  I'm
> not completely sure it will be harmless, so it won't be fixed in 24.1.

Ok.


Regards,

Michael.

bug#11258: 24.1.50; lexical binding inconsistencies

[Michael Heerdegen]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

> I think the compiler is right, so the interpreter needs fixing.  I'm not
> completely sure it will be harmless, so it won't be fixed in 24.1.

Testing my recipe today with (emacs-version) => "GNU Emacs 24.3.50.1
(x86_64-pc-linux-gnu, GTK+ Version 3.4.2)\n of 2013-04-10 on dex,
modified by Debian", I see that the behavior is still the same.


Regards,

Michael.

bug#11258: 24.1.50; lexical binding inconsistencies

[Stefan Monnier]

> Testing my recipe today with (emacs-version) => "GNU Emacs 24.3.50.1
> (x86_64-pc-linux-gnu, GTK+ Version 3.4.2)\n of 2013-04-10 on dex,
> modified by Debian", I see that the behavior is still the same.

I installed the trivial patch below, which should fix it.  Sorry for
that ridiculous delay.


        Stefan


=== modified file 'src/eval.c'
--- src/eval.c	2013-08-12 07:12:07 +0000
+++ src/eval.c	2013-08-28 18:24:19 +0000
@@ -2149,6 +2149,8 @@
   if (SYMBOLP (fun) && !NILP (fun)
       && (fun = XSYMBOL (fun)->function, SYMBOLP (fun)))
     fun = indirect_function (fun);
+  else
+    fun = Ffunction (Fcons (fun, Qnil));
 
   if (SUBRP (fun))
     {