post-command-hook

post-command-hook

[martin rudalics]

After updating my trunk I'm completely lost with the following issue:
With emacs -Q, switch to the buffer *Messages*, make it empty and
evaluate the following forms in it:

(defun foo ()
   (message "foo"))

(defun bar ()
   (message "bar"))

(add-hook 'post-command-hook 'foo)
(add-hook 'post-command-hook 'bar nil t)

Now move around that buffer.  Here it only runs the local hook function
"bar".  It does _not_ run the global hook function "foo".

All my earlier Emacs versions run both hooks alternately.

GNU Emacs 24.0.50.1 (i386-mingw-nt5.1.2600)
  of 2011-04-01 on NESTOR

Help appreciated, martin

Re: post-command-hook

[Antoine Levitt]

01/04/11 17:58, martin rudalics
> After updating my trunk I'm completely lost with the following issue:
> With emacs -Q, switch to the buffer *Messages*, make it empty and
> evaluate the following forms in it:
>
> (defun foo ()
>   (message "foo"))
>
> (defun bar ()
>   (message "bar"))
>
> (add-hook 'post-command-hook 'foo)
> (add-hook 'post-command-hook 'bar nil t)
>
> Now move around that buffer.  Here it only runs the local hook function
> "bar".  It does _not_ run the global hook function "foo".
>
> All my earlier Emacs versions run both hooks alternately.
>
> GNU Emacs 24.0.50.1 (i386-mingw-nt5.1.2600)
>  of 2011-04-01 on NESTOR
>
> Help appreciated, martin

Ah, this is probably related to the bug I'm seeing
(http://permalink.gmane.org/gmane.emacs.devel/137894). This setup should
be easier to debug though.

I think buffer-local post-command-hooks are messed up somehow (the other
code that triggers the bug also uses local hooks), but I can't figure
out the part that introduced it in
5e4cb836580a8ef2f9ea9b4b8317005941920027

Re: post-command-hook

[martin rudalics]

 > Ah, this is probably related to the bug I'm seeing
 > (http://permalink.gmane.org/gmane.emacs.devel/137894). This setup should
 > be easier to debug though.
 >
 > I think buffer-local post-command-hooks are messed up somehow (the other
 > code that triggers the bug also uses local hooks), but I can't figure
 > out the part that introduced it in

Juanma and me suppose it's revno:103769
(revid:monnier@iro.umontreal.ca-20110328202635-2t9ij68etiasfyy2).
Somewhere in run_hook_with_args, I suppose.  Unfortunately, that change
also involves some code cleanup so I was not able so far to track this
down.  Moreover it makes working with the trunk currently impossible for
me :-(

martin

Re: post-command-hook

[Antoine Levitt]

02/04/11 09:02, martin rudalics
>> Ah, this is probably related to the bug I'm seeing
>> (http://permalink.gmane.org/gmane.emacs.devel/137894). This setup should
>> be easier to debug though.
>>
>> I think buffer-local post-command-hooks are messed up somehow (the other
>> code that triggers the bug also uses local hooks), but I can't figure
>> out the part that introduced it in
>
> Juanma and me suppose it's revno:103769
> (revid:monnier@iro.umontreal.ca-20110328202635-2t9ij68etiasfyy2).

Yup, that's what I thought too.

> Somewhere in run_hook_with_args, I suppose.  Unfortunately, that change
> also involves some code cleanup so I was not able so far to track this
> down.  Moreover it makes working with the trunk currently impossible for
> me :-(

Stefan should take a look, but I guess he's busy with the lexbind merge
right now :) I'm simply using an old revision for the time being.

Re: post-command-hook

[Juanma Barranquero]

On Fri, Apr 1, 2011 at 17:58, martin rudalics <rudalics@gmx.at> wrote:

> (defun foo ()
>  (message "foo"))

If you change that to

(defun foo ()
  (message "foo")
  nil)

it works, so it seems like this part of the change is the culprit:


@@ -2543,9 +2577,7 @@
       GCPRO3 (sym, val, global_vals);

       for (;
-	   CONSP (val) && ((cond == to_completion)
-			   || (cond == until_success ? NILP (ret)
-			       : !NILP (ret)));
+	   CONSP (val) && NILP (ret);
 	   val = XCDR (val))
 	{
 	  if (EQ (XCAR (val), Qt))


    Juanma

Re: post-command-hook

[Stefan Monnier]

> (defun foo ()
>   (message "foo"))
> (defun bar ()
>   (message "bar"))
> (add-hook 'post-command-hook 'foo)
> (add-hook 'post-command-hook 'bar nil t)

I think the patch below fixes it,


        Stefan


=== modified file 'src/keyboard.c'
--- src/keyboard.c	2011-04-02 18:34:05 +0000
+++ src/keyboard.c	2011-04-02 21:37:19 +0000
@@ -1815,7 +1815,8 @@
 safe_run_hooks_1 (void)
 {
   eassert (CONSP (Vinhibit_quit));
-  return call0 (XCDR (Vinhibit_quit));
+  call0 (XCDR (Vinhibit_quit));
+  return Qnil;
 }
 
 /* Subroutine for safe_run_hooks: handle an error by clearing out the function

Re: post-command-hook

[Antoine Levitt]

02/04/11 23:40, Stefan Monnier
>> (defun foo ()
>>   (message "foo"))
>> (defun bar ()
>>   (message "bar"))
>> (add-hook 'post-command-hook 'foo)
>> (add-hook 'post-command-hook 'bar nil t)
>
> I think the patch below fixes it,

It fixes my problems, thanks!

Re: post-command-hook

[Johan Bockgård]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> (defun foo ()
>>   (message "foo"))
>> (defun bar ()
>>   (message "bar"))
>> (add-hook 'post-command-hook 'foo)
>> (add-hook 'post-command-hook 'bar nil t)
>
> I think the patch below fixes it,
>

>  safe_run_hooks_1 (void)
>  {
>    eassert (CONSP (Vinhibit_quit));
> -  return call0 (XCDR (Vinhibit_quit));
> +  call0 (XCDR (Vinhibit_quit));
> +  return Qnil;
>  }

Shouldn't the "return nil" be in safe_run_hook_funcall instead?

Re: post-command-hook

[martin rudalics]

> I think the patch below fixes it,

It fixes my problem.  Thanks.

martin

Re: post-command-hook

[Stefan Monnier]

> Shouldn't the "return nil" be in safe_run_hook_funcall instead?

Indeed, thanks for caching it before I committed it,


        Stefan