Error on leaving Gnus group

Error on leaving Gnus group

[Chong Yidong]

When I type "q" in the Gnus summary buffer to leave the group, I now
obtain the following backtrace:

Debugger entered--Lisp error: (error "No 1.0 specs in (vertical 1 (group 1 point))")
  signal(error ("No 1.0 specs in (vertical 1 (group 1 point))"))
  error("No 1.0 specs in %s" (vertical 1 (group 1 point)))
  gnus-configure-frame((vertical 1.0 (group 1 point)))
  gnus-configure-windows(group force)
  gnus-summary-exit()
  call-interactively(gnus-summary-exit nil nil)

This problem was introduced by this change:


2011-01-28  Julien Danjou  <julien@danjou.info>

	* gnus-win.el: Remove dead function gnus-window-configuration-element.
	(gnus-all-windows-visible-p): Remove old compatibility code.
	(gnus-window-top-edge): Add docstring.

If I put the "old compatibility code" back into
gnus-all-windows-visible-p, the error is not signalled.  Dunno what the
"old compatibility" refers to; I don't have much in the way of Gnus
customization, and using a fresh .newsrc.eld file gives the same error.

Does this ring a bell for anyone?  If not, I can try making a test case.

Re: Error on leaving Gnus group

[Lars Ingebrigtsen]

Chong Yidong <cyd@stupidchicken.com> writes:

> When I type "q" in the Gnus summary buffer to leave the group, I now
> obtain the following backtrace:
>
> Debugger entered--Lisp error: (error "No 1.0 specs in (vertical 1 (group 1 point))")

[...]

> If I put the "old compatibility code" back into
> gnus-all-windows-visible-p, the error is not signalled.  Dunno what the
> "old compatibility" refers to;

I vaguely remember the window spec thing being different way back when.
"1" now means "use 1 line (or column)", while "1.0" means "use the
entire window".

If anybody still has the old-style window conf spec, then the old compat
code has to be put back in.

> I don't have much in the way of Gnus customization, and using a fresh
> .newsrc.eld file gives the same error.

You don't have any code anywhere that says "vertical 1" somewhere?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Error on leaving Gnus group

[Julien Danjou]

[-- Attachment #1: Type: text/plain, Size: 268 bytes --]

On Mon, Feb 07 2011, Lars Ingebrigtsen wrote:

> If anybody still has the old-style window conf spec, then the old compat
> code has to be put back in.

The code I deleted was marked as deprecated since 1998.

-- 
Julien Danjou
❱ http://julien.danjou.info

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: Error on leaving Gnus group

[Chong Yidong]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> I vaguely remember the window spec thing being different way back when.
> "1" now means "use 1 line (or column)", while "1.0" means "use the
> entire window".
>
> If anybody still has the old-style window conf spec, then the old compat
> code has to be put back in.
>
>> I don't have much in the way of Gnus customization, and using a fresh
>> .newsrc.eld file gives the same error.
>
> You don't have any code anywhere that says "vertical 1" somewhere?

Nope.  Here's my recipe:

0. Make a .emacs file with only the following line:

  (setq gnus-agent nil gnus-select-method '(nntp "news.gmane.org"))

1. emacs
2. M-x gnus RET
3. A A    (to list the gmane groups)
4. SPC    (to select gmane.announce)
5. 3 RET  (to grab 3 articles)
6. q

  Error ==> No 1.0 specs in (vertical 1 (group 1 point))

Re: Error on leaving Gnus group

[Julien Danjou]

[-- Attachment #1: Type: text/plain, Size: 2191 bytes --]

On Sun, Feb 06 2011, Chong Yidong wrote:

> Debugger entered--Lisp error: (error "No 1.0 specs in (vertical 1 (group 1 point))")
>   signal(error ("No 1.0 specs in (vertical 1 (group 1 point))"))
>   error("No 1.0 specs in %s" (vertical 1 (group 1 point)))
>   gnus-configure-frame((vertical 1.0 (group 1 point)))
>   gnus-configure-windows(group force)
>   gnus-summary-exit()
>   call-interactively(gnus-summary-exit nil nil)
>
> This problem was introduced by this change:
>
>
> 2011-01-28  Julien Danjou  <julien@danjou.info>
>
> 	* gnus-win.el: Remove dead function gnus-window-configuration-element.
> 	(gnus-all-windows-visible-p): Remove old compatibility code.
> 	(gnus-window-top-edge): Add docstring.
>
> If I put the "old compatibility code" back into
> gnus-all-windows-visible-p, the error is not signalled.  Dunno what the
> "old compatibility" refers to; I don't have much in the way of Gnus
> customization, and using a fresh .newsrc.eld file gives the same error.
>
> Does this ring a bell for anyone?  If not, I can try making a test case.

This is real diff behind that change:
--- a/lisp/gnus-win.el
+++ b/lisp/gnus-win.el
@@ -437,11 +437,7 @@ should have point."
        type buffer win buf)
     (while (and (setq split (pop stack))
                all-visible)
-      ;; Be backwards compatible.
-      (when (vectorp split)
-       (setq split (append split nil)))
-      (when (or (consp (car split))
-               (vectorp (car split)))
+      (when (consp (car split))
        (push 1.0 split)
        (push 'vertical split))
       ;; The SPLIT might be something that is to be evaled to

I do not think it's wrong. There's no vector involed here.

In your backtrace I see
   gnus-configure-frame((vertical 1.0 (group 1 point)))

I cannot find such a line in Gnus code source.

In gnus-win there something that looks like it which is:
  '((group
     (vertical 1.0
	       (group 1.0 point)))

But it's 1 dot zero.

So I don't get where it come from. Maybe you have an very definition of
the gnus-buffer-configuration variable? If that's possible.

-- 
Julien Danjou
❱ http://julien.danjou.info

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: Error on leaving Gnus group

[Lars Ingebrigtsen]

Julien Danjou <julien@danjou.info> writes:

> In gnus-win there something that looks like it which is:
>   '((group
>      (vertical 1.0
> 	       (group 1.0 point)))
>
> But it's 1 dot zero.
>
> So I don't get where it come from. Maybe you have an very definition of
> the gnus-buffer-configuration variable? If that's possible.

Was this resolved?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Error on leaving Gnus group

[Chong Yidong]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Was this resolved?

I'm still seeing it, even with an empty .emacs and a fresh
.newsrc.eld...

Debugger entered--Lisp error: (error "No 1.0 specs in (vertical 1 (group
  1 point))")
  signal(error ("No 1.0 specs in (vertical 1 (group 1 point))"))
  error("No 1.0 specs in %s" (vertical 1 (group 1 point)))
  gnus-configure-frame((vertical 1.0 (group 1 point)))
  gnus-configure-windows(group force)
  gnus-summary-exit()
  call-interactively(gnus-summary-exit nil nil)

Attempting to edebug this, however, makes it disappear.  One possibility
I can think of is that the (equal size 1.0) at gnus-win.el:316 is
triggering a floating point error somewhere in the byte compiler.  I'll
play around with a it a little more.

Re: Error on leaving Gnus group

[Chong Yidong]

Lars Ingebrigtsen <larsi@gnus.org> writes:

>> In gnus-win there something that looks like it which is:
>>   '((group
>>      (vertical 1.0
>> 	       (group 1.0 point)))
>>
>> But it's 1 dot zero.
>
> Was this resolved?

OK, I can confirm that if I replace the line

   (cond ((equal size 1.0)

in gnus-win.el:316 with

   (cond ((= size 1.0)

then the problem goes away.  Still not sure where the integer to float
conversion is taking place, but my guess is somewhere in the byte
optimizer, because if I delete gnus-win.elc, the problem goes away too.

So this looks like a bug in Emacs itself.