bug#5405: select-frame losing current-buffer

bug#5405: select-frame losing current-buffer

[Uday S Reddy]

The documentation of make-frame says that current-buffer continues to
selected in the new frame.  The documentation of select-frame doesn't
say anything about the matter, but one would normally expect that the
current-buffer should still remain the same.  However, the following
example shows that select-frame loses the current-buffer:

(defun testing ()
  (interactive)
  (let ((new-buffer (get-buffer-create " testing")))
    (set-buffer new-buffer)
    (select-frame (make-frame nil))
    (if (not (equal (current-buffer) new-buffer))
	(debug))))

I presume that the space at the beginning of the buffer name is
a partial cause of this misbehaviour.

Cheers,
Uday Reddy



In GNU Emacs 22.3.1 (i386-mingw-nt5.1.2600)
 of 2008-09-06 on SOFT-MJASON
Windowing system distributor `Microsoft Corp.', version 5.1.2600
configured using `configure --with-gcc (3.4)'

Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: nil
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: ENU
  locale-coding-system: cp1252
  default-enable-multibyte-characters: t

Major mode: Mail

Minor modes in effect:
  savehist-mode: t
  encoded-kbd-mode: t
  tooltip-mode: t
  mouse-wheel-mode: t
  menu-bar-mode: t
  file-name-shadow-mode: t
  global-font-lock-mode: t
  font-lock-mode: t
  blink-cursor-mode: t
  unify-8859-on-encoding-mode: t
  utf-translate-cjk-mode: t
  line-number-mode: t
  transient-mark-mode: t
  abbrev-mode: t

Recent input:
C-x b * s c SPC <return> <f1> C-x . <return> ( d e 
f u n SPC t e s t i n g SPC ( ) <return> TAB C-( ( 
i n t e r a c t i v e ) <return> TAB ( l e t SPC ( 
( n e w - b u f f e r SPC ( g e t - b u f f e r - c 
r e a t e SPC " SPC t e s t i n g " ) ) ) <return> 
TAB ( s e t - b u f f e r SPC n e w - b u f f e r ) 
<return> TAB ( l e t SPC ( ( n e w - f r a m e SPC 
C-a C-k TAB ( s e l e c t - f r a m e SPC ( m a k e 
- f r a m e SPC n i l ) ) <return> TAB ( i f SPC ( 
n o t SPC ( e u q <backspace> <backspace> q u a l SPC 
( c u r r e n t - b u f f e r ) SPC n e w - b u f f 
e r ) ) <return> TAB 9 d e <backspace> <backspace> 
<backspace> ( d e b u g ) ) ) ) <escape> C-x M-x t 
e s t i n g <return> q C-x b <return> C-x b * s c SPC 
<return> C-p C-p C-p C-p C-p C-p C-a C-SPC C-n C-n 
C-n C-n C-n C-n C-n <return> C-x C-x C-x C-k C-x m 
M-x r e p o r t - e m SPC SPC <return>

Recent messages:
There are unsaved changes, quit anyway? (y or n) 
Delete auto save file y:/tmpmail/#vm-bug-forwarding.msg#? (y or n) 
Mark set
Auto-saving...done
testing
Entering debugger...
Back to top level.
Mark set
Loading mail-utils...done
Loading emacsbug...done

bug#5405: select-frame losing current-buffer

[Chong Yidong]

Uday S Reddy <u.s.reddy@cs.bham.ac.uk> writes:

> The documentation of make-frame says that current-buffer continues to
> selected in the new frame.  The documentation of select-frame doesn't
> say anything about the matter, but one would normally expect that the
> current-buffer should still remain the same.  However, the following
> example shows that select-frame loses the current-buffer:
>
> (defun testing ()
>   (interactive)
>   (let ((new-buffer (get-buffer-create " testing")))
>     (set-buffer new-buffer)
>     (select-frame (make-frame nil))
>     (if (not (equal (current-buffer) new-buffer))
> 	(debug))))
>
> I presume that the space at the beginning of the buffer name is
> a partial cause of this misbehaviour.

This is deliberate behavior dating back about a decade (frame.c:392).
Buffers whose names start with a space are considered "hidden buffers"
that should not ordinarily be displayed (e.g. they don't show up in
M-x list-buffers either).  I'll update the documentation to mention
this.

bug#5405: select-frame losing current-buffer

[Uday S Reddy]

Dear Chong,

Thanks very much for the quick response.  But I am not convinced the
hidden buffer idea explains the behaviour I found.  Before doing
select-frame, the "hidden buffer" is the current buffer.  For whatever
reason, the user or the code chose it as the current buffer.  I don't
believe that the buffer should be forcibly dumped and the focus placed
on some other random buffer that happens to be around.

This behaviour was found in maintaining VM which, for some indepdent
reasons, chose a buffer name with a leading space, and some serious
buffer corruption resulted from it.  This seems dangerous, undesirable
behaviour. 

I have now modified VM to avoid the leading space.  So the issue
doesn't affect me any more.  But it took me a day's labour to find the
problem.  I hope there won't be others who will get simiarly burned.

Cheers,
Uday



Chong Yidong writes:

> Uday S Reddy <u.s.reddy@cs.bham.ac.uk> writes:
> 
> > The documentation of make-frame says that current-buffer continues to
> > selected in the new frame.  The documentation of select-frame doesn't
> > say anything about the matter, but one would normally expect that the
> > current-buffer should still remain the same.  However, the following
> > example shows that select-frame loses the current-buffer:
> >
> > (defun testing ()
> >   (interactive)
> >   (let ((new-buffer (get-buffer-create " testing")))
> >     (set-buffer new-buffer)
> >     (select-frame (make-frame nil))
> >     (if (not (equal (current-buffer) new-buffer))
> > 	(debug))))
> >
> > I presume that the space at the beginning of the buffer name is
> > a partial cause of this misbehaviour.
> 
> This is deliberate behavior dating back about a decade (frame.c:392).
> Buffers whose names start with a space are considered "hidden buffers"
> that should not ordinarily be displayed (e.g. they don't show up in
> M-x list-buffers either).  I'll update the documentation to mention
> this.
> 
> 

bug#5405: select-frame losing current-buffer

[Drew Adams]

> > > The documentation of make-frame says that current-buffer 
> > > continues to selected in the new frame.  The
> > > documentation of select-frame doesn't
> > > say anything about the matter, but one would normally 
> > > expect that the current-buffer should still remain the same.  
...
> > > I presume that the space at the beginning of the buffer name is
> > > a partial cause of this misbehaviour.
> > 
> > This is deliberate behavior dating back about a decade 
> > (frame.c:392). Buffers whose names start with a space are
> > considered "hidden buffers" that should not ordinarily be
> > displayed (e.g. they don't show up in M-x list-buffers either).  
> > I'll update the documentation to mention this.
>
> Thanks very much for the quick response.  But I am not convinced the
> hidden buffer idea explains the behaviour I found.  Before doing
> select-frame, the "hidden buffer" is the current buffer.  For whatever
> reason, the user or the code chose it as the current buffer.  I don't
> believe that the buffer should be forcibly dumped and the focus placed
> on some other random buffer that happens to be around.
> 
> This behaviour was found in maintaining VM which, for some indepdent
> reasons, chose a buffer name with a leading space, and some serious
> buffer corruption resulted from it.  This seems dangerous, undesirable
> behaviour. 
> 
> I have now modified VM to avoid the leading space.  So the issue
> doesn't affect me any more.  But it took me a day's labour to find the
> problem.  I hope there won't be others who will get simiarly burned.

Without speaking directly to what `select-frame'/`make-frame' should do in this
regard, let me say that speaking in terms of "normally expect" (Uday) and
"should not ordinarily" (Yidong) simply begs the question. No such general rule
can answer the question completely for this particular situation.

By _default_, in many common situations, such buffers are hidden or ignored in
some _particular sense_ - for example, as completion candidates. That does not
mean that they must or should be ignored in all contexts, or that users should
not have a way to override a default behavior that ignores them.

The reason for such a general/default/common treatment is more important than
the "rule" itself, as only it provides guidance. The reason is that not ignoring
such buffers can distract or confuse users. Users _typically_ do not need to
consider such buffers - they just get in the way. But the fact that users _can_
consider them if they want (including for completion) speaks to not casting such
a generally helpful rule in concrete.

IOW, the question was raised - for `select-frame'/`make-frame' in particular -
and it is still raised (unanswered). The argument that such buffers must by
definition be ignored is not a valid one. It begs the question to be answered -
both generally, and in this particular case.

Wrt `select-frame'/`make-frame': I'm not sure what the right behavior is. If, as
Uday says, the "hidden" buffer was already chosen then, analogously to
completion when the user explicitly types a pattern that chooses an otherwise
hidden buffer, that choice should probably be respected. If the user types
`foo.log' or ` *foo*' during completion, we don't refuse to give access to that
buffer under the pretext that it is a hidden buffer.

If a program or user has already chosen, then we should probably let it be. But
there might be other considerations for this particular case, which would also
need to be taken into account - dunno. Only attention to the details can help,
not some general rule about a priori hiddenness.

bug#5405: select-frame losing current-buffer

[martin rudalics]

 > The documentation of make-frame says that current-buffer continues to
 > selected in the new frame.

The documentation also says that

   The previously selected frame remains selected.  However, the
   window system may select the new frame for its own reasons, for
   instance if the frame appears under the mouse pointer and your
   setup is for focus to follow the pointer.

so the current buffer may change (as described below) in some cases.

 > The documentation of select-frame doesn't
 > say anything about the matter, but one would normally expect that the
 > current-buffer should still remain the same.

The documentation of `select-frame' says

   Subsequent editing commands apply to its selected window.

which inherently means that that frame's selected window becomes the
selected window.  Selecting a window WINDOW means to assure that

   If WINDOW is not already selected, make WINDOW's buffer current
   and make WINDOW the frame's selected window.

Which doesn't mean that the documentation can't be improved ;-)

 > I presume that the space at the beginning of the buffer name is
 > a partial cause of this misbehaviour.

It's not related.

martin

bug#5405: select-frame losing current-buffer

[Uday S Reddy]

I didn't understand this message entirely, but I can confirm that the
problem arises in interactive use as well.  If you have the focus on a
non-hidden buffer and create a new-frame, then this buffer is
displayed in the new frame.  If it is a "hidden" buffer then it is
ignored and some other buffer is displayed in the new frame.  This can
be confusing and perhaps annoying to the users, but it is probably not
a serious problem during interactive use.

Buf if the same problem occurs inside application code, then it can be
more dangerous because the code might errorneously assume that the
correct buffer is in place and modify it.  The fact that it is a
*special case* means that it might go unnoticed during testing and
faulty application code gets delivered.  

So, I recommend that no special case should be made for "hidden"
buffers.  As Drew has argued cogently, the case for the special case
is weak.

Cheers,
Uday




martin rudalics writes:

>  > The documentation of make-frame says that current-buffer continues to
>  > selected in the new frame.
> 
> The documentation also says that
> 
>    The previously selected frame remains selected.  However, the
>    window system may select the new frame for its own reasons, for
>    instance if the frame appears under the mouse pointer and your
>    setup is for focus to follow the pointer.
> 
> so the current buffer may change (as described below) in some cases.
> 
>  > The documentation of select-frame doesn't
>  > say anything about the matter, but one would normally expect that the
>  > current-buffer should still remain the same.
> 
> The documentation of `select-frame' says
> 
>    Subsequent editing commands apply to its selected window.
> 
> which inherently means that that frame's selected window becomes the
> selected window.  Selecting a window WINDOW means to assure that
> 
>    If WINDOW is not already selected, make WINDOW's buffer current
>    and make WINDOW the frame's selected window.
> 
> Which doesn't mean that the documentation can't be improved ;-)
> 
>  > I presume that the space at the beginning of the buffer name is
>  > a partial cause of this misbehaviour.
> 
> It's not related.
> 
> martin
> 

bug#5405: select-frame losing current-buffer

[Lars Magne Ingebrigtsen]

Uday S Reddy <u.s.reddy@cs.bham.ac.uk> writes:

> So, I recommend that no special case should be made for "hidden"
> buffers. 

I think this is a feature, and not a bug, so I'm closing this report.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/