Re: Small change to _NET_CURRENT_DESKTOP messages?



On Mon, 28 Feb 2005 19:11:57 +0100, Lubos Lunak <l lunak suse cz> wrote:
> On Wednesday 16 of February 2005 20:25, Elijah Newren wrote:
> > Hi,
> >
> > When we receive _NET_CURRENT_DESKTOP messages, we are picking a window
> > to focus.  We'd like to use a real timestamp for that instead of
> > making one up.  Would it be okay to add a timestamp field to
> > _NET_CURRENT_DESKTOP messages in the EWMH?  Simple one-line patch that
> > does so is attached...
> 
>  It should be hopefully fine. However there probably should be a note stating
> that with older apps the timestamp may be zero (which is obvious I guess),

I attached a new version of the patch that does this.  Okay to commit?

> and the less obvious that with even older apps the field may be
> uninitialized. It wasn't _that_ long ago when I added all those "other
> data.l[] elements = 0", and I think at least the GNOME implementation hadn't
> been doing this before this change.
> 
>  *checking*
> 
>  Slightly less than two years ago, actually. I wonder if we can start more or
> less ignoring such old implementations by now.

Well, we've done so for things like _NET_ACTIVE_WINDOW, right?  If
we're really worried about this, then I think we'd have to rename
_NET_ACTIVE_WINDOW and anything else that has had any fields added. 
Personally, given that the _NET_ACTIVE_WINDOW addition seems to be
fine, I think we should ignore these older implementations, just as we
have to ignore buggy clients that break the spec.

Elijah
Index: wm-spec.xml
===================================================================
RCS file: /cvs/icccm-extensions/wm-spec/wm-spec.xml,v
retrieving revision 1.15
diff -p -u -r1.15 wm-spec.xml
--- wm-spec.xml	25 Jan 2005 16:55:05 -0000	1.15
+++ wm-spec.xml	28 Feb 2005 18:27:45 -0000
@@ -395,8 +395,13 @@ _NET_CURRENT_DESKTOP
   message_type = _NET_CURRENT_DESKTOP 
   format = 32
   data.l[0] = new_index
+  data.l[1] = timestamp
   other data.l[] elements = 0
 ]]></programlisting>
+		<para>
+Note that the timestamp may be 0 for clients using an older version of
+this spec, in which case the timestamp field should be ignored.
+		</para>
 	</sect2><sect2><title>_NET_DESKTOP_NAMES</title>
 	<programlisting><![CDATA[
 _NET_DESKTOP_NAMES, UTF8_STRING[]


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]