Re: [dwm] remembering client order in each tag

From: Anthony Brown <>
Date: Thu, 7 Sep 2006 07:53:07 +0200

On Wed, Sep 06, 2006 at 08:23:54PM +0200, Sander van Dijk wrote:
> On 9/6/06, Anthony Brown <> wrote:
> >The behaviour for the latest changeset (442) is, as I described before,
> >basically they way I would expect (i.e. like) things to behave. I have
> >now also verified that z-ordering for clients with a certain tag is also
> >preserved after doing a view-all and switching back to viewing a single
> >tag. The only slight problem remaining is the mod-tab behaviour. However
> >this is only a minor irritant.
> It's not just a minor irritant, it means that keyboard based focusing
> is fundamentally broken now. Focus changes should _never_ have an
> effect on client ordering. Currently, nextfocusing in floating mode
> loops between the first two clients, which is a severe conceptual
> error.
> >The reason I like to have the z-order preservation is for working with
> >floating clients that overlap. This happens for example when I work with
> >an editor in combination with a graphing programme. It is no good to
> >have plot windows managed in tiled mode as they should be of a certain
> >size and shape to make the graphics readable. At the same time they may
> >overlap with the editor and loosing the z-ordering when temporarily
> >switching to xterms with a different tag becomes annoying after a while.
> Like I said, z-ordering _is_ preserved in 1.4, it's just that when you
> switch to a view that doesn't contain the currently focused client,
> dwm will focus and _raise_ the newly focused window. So far, this is
> the _only_ place where z-order is (non-manually) altered. You can
> verify this by opening say 10 clients in a view and then switch
> between tags. You'll see that the newly focused window is brought to
> the top, but that _all_ the other windows retain their z-order.
> Basically, the introduction of the nextfocusing problem indicates
> mainly this: we're trying to fix something in the _wrong_place_. The
> situation should be reverted to what it was like in 1.4, and dwm
> should do something like this instead when switching to a view that
> does not have the currenly selected client in it:
> IF there's at least one tiled client
> THEN focus the mastertile
> ELSE if there are any floating clients
> THEN focus the topmost window (rather than the first client in the
> clientlist)
> ELSE put focus on the root window
> That would fix both focus cycling _and_ your issue, and
> _the_right_way_, I might add.
> Sorry for sounding a little harsh, but I get a little prickly when
> people call a severe conceptual mistake "just a minor irritant".
> Client (focus) order and current z-order are _completely_unrelated_
> and like a said earlier in this thread, tying them together is bound
> to break at least one of the two...
> Just to be clear: I'm trying to keep dwm from going the wrong way
> here, not to hurt your feelings. My apologies if I did.
> Greetings, Sander.

Don't worry about my feelings Sander, I can take a bit of criticism ;)

In any case, there seems to be a misunderstanding about the behaviour
that I would like to see. I don't so much care about preserving the
*focus* when switching between tags (in tiled mode this will always
revert to the master column anyway) but the *z-ordering*. That is, when
in float mode I decided on a certain ordering of overlapping clients,
for whatever reason, this should be left untouched by the window
manager. I highly appreciate the fact that in tiled mode DWM does all
the laying out for me, however in float mode I would like to stay in
control. So again, it is only the z-ordering that concerns me, not the
focus. Thus decouping these two issues is fine with me.

I agree that the mod-tab cycling should work properly.

Received on Thu Sep 07 2006 - 07:53:14 UTC

This archive was generated by hypermail 2.2.0 : Sun Jul 13 2008 - 14:31:02 UTC