The "Close" option in the tasklist menu does the same thing.
Is this option not there ? or am I getting your point wrong ?

There used to be "close" and "kill"

"kill" disconnected the program from the X server, which would sometimes 
cause the program to exit (and sometimes cause the program 
to continue in its infinite loop or whatever was wrong with it).

"close" will now offer to kill the process of (vs. just disconnect)
unresponsive apps, for GTK 2 apps that support detection of 
non-responsiveness.

it's impossible to explain the difference between close and kill, 
in fact most people who used kill did not understand what it did
(just disconnection, rather than actual kill), so having both 
in the UI was a terrible idea.

I actually have a patch to add a "Kill Application" action button ...
its a more friendly version of xkill. (Calum wants it renamed "Force
Quit" or something silly like that).

I think it makes sense to have this since we can't completely rely on
_NET_CLOSE_WINDOW to actually kill an app.

Re-opening so I don't forget to get the patch in order.


=====================
o Sends a _NET_CLOSE_WINDOW ClientMessage to the root window
  indicating the window to close.
o According to the spec, window managers may try and be more
  clever than simply trying to close the window.
o In response to this message Metacity will do the following

    if (supports_wm_protocol (window, WM_DELETE_WINDOW)
      send_wm_delete_window_message (window)
    else
      XKillClient (window);

    if (supports_wm_protocol (window, _NET_WM_PING))
      if (send_net_wm_ping_message (window) == no_reply)
        if (pop_up_delete_dialog_question (window) == delete)
          {
            if (on_same_host (window) && has_pid_prop_set (window))
              kill (window->pid, SIGKILL)
            XKillClient (window);
          }

o Clearly doesn't work for us in the case where we have a hung app
  which supports WM_DELETE_WINDOW, but doesn't support _NET_WM_PING.

you can't rely on XKillClient() to kill an app either, though.
In fact the most recent cases of hung clients I can remember 
(Netscape 4's love of infinite loops, and XChat hanging on a thread
lock) would not be solved by XKillClient(). "Force quit" is kind of 
a misnomer here. More like "force to become invisible yet still use 
my CPU"

But you know. "Force quit button" is I guess mostly harmless, as long
as it isn't in the window list.

This could confuse users, tho, yes?  Why give them a button that
doesn't do as "advertised" (to their minds) vs. finding a better
method/procedure first that might do the job all the way?

Perhaps poll the app and see if it's shutting down yet after X second
of XKillClient or something?

Fixed on HEAD:

2003-01-24  Mark McLoughlin  <mark@skynet.ie>
                                                                     
                                         
        * Makefile.am: add panel-force-quit.[ch] to the build.
                                                                     
                                         
        * panel-force-quit.[ch]: impl a safer more GNOMEish version
        of xkill.