bug#19413: 24.4; (desktop-save-mode 1) save frame as page and desktop sticky (_NET_WM_STATE_STICKY) when close Emacs shaded, (setq desktop-restore-frames nil) fix issue.

[Oleksandr Gavenko <gavenkoa@gmail.com>]

On 2014-12-20, martin rudalics wrote:

>> When frame sticky (Fvwm decorate it especially) I get "t". For non-sticky I
>> get "nil".
>
> So the value of sticky corresponds to what you see on screen.  Correct?

Yes.

> To get it right: You first close Emacs with a shaded frame and if you
> afterwards look at the desktop file it has (sticky) in it.  If you now
> restart Emacs and close it again, the desktop file has (sticky . t).

Yes.

>> So there are two problems:
>>
>>   1) Emacs instead saving shaded window attribute add "(sticky)" to ~/.emacs.desktop
>
> So if you do not have a desktop file initially but enable
> `desktop-save-mode' and then
>
> (1) close Emacs with a shaded frame, and
>
> (2) close Emacs with an unshaded frame
>
> the difference is a (sticky) entry for (1) and no such entry for (2).
> Is that correct?

Yes.

>>   2) (desktop-read) wrongly apply (sticky), and (sticky . t) and (sticky . nil)
>>   - as always sticky.
>
> This sounds like a reasonable explanation.  x_net_wm_state has
>
> /**   store_frame_param (f, Qsticky, sticky ? Qt : Qnil); **/
>
> without any further comment, maybe it's related.  Does Emacs process
>
> (set-frame-parameter nil 'sticky ...)
>
> correctly on your system without desktop saving involved?

Any of:

  (set-frame-parameter nil 'sticky t)
  (set-frame-parameter nil 'sticky nil)
  (set-frame-parameter nil 'sticky 1)
  (set-frame-parameter nil 'sticky -1)

make clean "emacs -Q" sticky. Seems that issue here.

It is possible Fvwm bug but I don't know how to distinct. In FvwmConsole I
run:

  WindowId 0x2400013 Stick false
  WindowId 0x2400013 Stick true

and Emacs window change stickness exactly to command.

>> I check /src/xterm.h:
>>
>>    /* Atoms dealing with EWMH (i.e. _NET_...) */
>>    Atom Xatom_net_wm_state, Xatom_net_wm_state_fullscreen,
>>      Xatom_net_wm_state_maximized_horz, Xatom_net_wm_state_maximized_vert,
>>      Xatom_net_wm_state_sticky, Xatom_net_wm_state_hidden,
>>      Xatom_net_frame_extents,
>>      Xatom_net_current_desktop, Xatom_net_workarea;
>>
>> There are no _NET_WM_STATE_SHADED that defined in:
>>
>>    http://standards.freedesktop.org/wm-spec/wm-spec-1.3.html#idm140130317598336
>>      _NET_WM_STATE
>>
>> which say:
>>
>>    Possible atoms are:
>>
>>    _NET_WM_STATE_MODAL, ATOM
>>    _NET_WM_STATE_STICKY, ATOM
>>    _NET_WM_STATE_MAXIMIZED_VERT, ATOM
>>    _NET_WM_STATE_MAXIMIZED_HORZ, ATOM
>>    _NET_WM_STATE_SHADED, ATOM
>>    _NET_WM_STATE_SKIP_TASKBAR, ATOM
>>    _NET_WM_STATE_SKIP_PAGER, ATOM
>>    _NET_WM_STATE_HIDDEN, ATOM
>>    _NET_WM_STATE_FULLSCREEN, ATOM
>>    _NET_WM_STATE_ABOVE, ATOM
>>    _NET_WM_STATE_BELOW, ATOM
>>    _NET_WM_STATE_DEMANDS_ATTENTION, ATOM
>
> We have to possibilities to deal with these: Either we wait for Jan to
> return from his voyage or you fix these issues yourself.  Given your
> findings above I'm quite confident that you could do it.

xterm.c and elisp frame handling code are complicate to me.

I just can summarise investigation: 

 * (set-frame-parameter nil 'sticky ...) doesn't work, it always set
   _NET_WM_STATE(ATOM) to _NET_WM_STATE_STICKY regardless passed value.

 * Emacs code base know nothing about _NET_WM_STATE_SHADED value of
   _NET_WM_STATE(ATOM).

 * There are no "shaded" (info "(elisp)Window Frame Parameters"). I see:
   ‘sticky’, ‘visibility’, ‘fullscreen’ (maximized, fullboth, fullwidth,
   fullheight).

 * Closing shaded Emacs with "(desktop-save-mode 1)" add (stick) parameter to
   ~/.emacs.desktop

-- 
Best regards!