Re: [STUMP] Interesting Bug...

[Ben Spencer]

On Fri, Jan 22, 2010 at 10:21:33PM -0500, Brit Butler wrote:
> 2) laptop enabled, VGA enabled with these results:
> address@hidden paktahn-dev]$ xdpyinfo -ext XINERAMA | grep head
>   head #0: 1680x1050 @ 0,0
>   head #1: 1280x800 @ 0,0
> STUMPWM> (screen-heads (current-screen))
> (#S(frame 0 NIL 0 0 1280 800))

This is a bug (or feature, depending on how you look at it).  When
there are overlapping heads, make-screen-heads removes duplicates
without looking at the size of the heads involved.  So in your case,
head #0 is eliminated just because it happens to be first and you get
a 1280x800 head.

What I'm not too sure about is what the correct behaviour should be
when this happens.  I can think of the following options:

1) Use the intersection of all the overlapping heads.  This would
result in wasted pixels on the larger ones.

2) Use the union of all overlapping heads.  I think this is a
non-starter because you could end up with sections that weren't
visible on any head.

3) Use the biggest head, relegating the smaller heads to second-class
citizens who can't see the whole desktop.

4) Subdivide the union of all the heads into visible rectangles and
treat those as 'heads'.  This would allow for effective use of every
visible bit of screen but is weird and unintuitive.

Of these, 2) and 3) would give you the behaviour you want, but both
would be less than ideal under other circumstances.


> 3) laptop disabled, VGA enabled with these results:
> address@hidden paktahn-dev]$ xdpyinfo -ext XINERAMA | grep head
>   head #0: 1680x1050 @ 0,0
> STUMPWM> (screen-heads (current-screen))
> (#S(frame 0 NIL 0 0 1680 1050))
> 
> The reason I switch completely, only running one display at a time, is
> because of my other problem. You'll notice (screen-heads (current-screen))
> remains at 1280x800 with the laptop + VGA monitor enabled. So enabling both
> "heads" only gets me one head listed for (screen-heads (current-screen))
> which fails to take advantage of the new screen space. This is why I wrote
> my fix-layout defcommand but that too only works when there is a single
> display enabled.

I'm still unclear on one point.  You say you have to disable the
laptop display for fix-layout to work, but when you do that stumpwm
should notice and resize the windows for you.  Are you saying you get
into a situation where (screen-heads (current-screen)) is (#S(frame 0
NIL 0 0 1680 1050)) (as above), but the windows are still 1280x800?

Did you have any luck with setting *top-level-error-action* to :break?
I haven't been able to reproduce your crash here so a backtrace would
be very useful for diagnosing it.


Ben