Re: knoppix 6.7 has many window focus problems
On Thu, Sep 15, 2011 at 09:15:34AM -0400, Andrew wrote:
> Dear Klaus,
> > Hello Andrew,
> > On Thu, Sep 15, 2011 at 02:35:00AM -0400, Andrew wrote:
> >> Dear Klaus,
> >> I have been observing these problems, including the lxterminal-logout
> >> window sometimes not getting focus, with metacity because I use the no3d
> >> cheat-code.
> > You mean "NOT" with metacity, right? metacity does not have the "prevent
> > focus stealing" feature that compiz uses.
> No, pid 3752 calls itself metacity:
> 3516 ? 00:00:00 knoppix-startx
> 3522 ? 00:00:00 su
> 3600 ? 00:00:00 startx
> 3619 ? 00:00:00 xinit
> 3620 tty5 00:11:38 Xorg
> 3628 ? 00:00:00 ck-launch-sessi
> 3734 ? 00:00:00 ssh-agent
> 3742 ? 00:00:00 lxsession
> 3752 ? 00:00:06 metacity
> 3755 ? 00:00:01 xscreensaver
> 3756 ? 00:00:12 lxpanel
> 3757 ? 00:00:00 pcmanfm
> 3764 ? 00:00:00 polkit-gnome-au
Ok, I see. I'm not sure abous metacitys window handling and session
management, have to investigate this by disabling compiz.
> >> As for the root terminal window with random dimensions, that is
> >> lxterminal. The same thing happens when running running a second (or
> >> third, or fourth, ...) window using ctrl+shift+N or typing lxterminal at
> >> the prompt. It happens for both root and for knoppix, with one
> >> exception:
> >> so far, every window I have opened with the quick launch tool bar shell
> >> icon has been 79x23.
> > I just noticed that the gksu desktop file (which is called "Root
> > Terminal" in the menu) startes with different options (transparent
> > background etc. in lxterminal) than our knoppix-special menu entry under
> > "Preferences" called "Root Shell". While it would probably be better to
> > only have ONE "Root" menu entry, I neither want to change the original
> > gksu setup, nor the knoppix-specific entry under "Preferences". Better
> > two working tools than (in the worst case) none.
> Oh wow! That's great! I recognize the command line from the Leonardo
> penguin Knoppix menu in the old quick launch tool bar! I would say
> "different options" is a bit of an understatement: It runs rxvt, not
> lxterminal. (By the way, rxvt consistently opens with 80x24.)
Right, the "root shell" terminal in Knoppix tries to use the simplest X
terminal application available, which is rxvt or xterm (the latter if
rxvt is not installed). These are not capable session management, i.e.
the window manager will just randomply place them on the screen without
remembering anything about their previous location or window size, and
just use whatever they have set as default.
> > The different dimensions when opening the window may be related to the
> > different settings in the specific tool.
> I don't understand. Personally, I suspect the randomly varying dimensions
> problem is probably in lxterminal. I note that a few knoppix versions
> ago, lxterminal used to open consistently at 80x24.
I did not change anything there. At least, I'm not aware of it. It may
just be an upgrade of lxsession that caused the change. :-/
> > I only feel responsible for the
> > "Root Shell" in "Preferences" because this is my own package, while
> > "gksu" is original Debian.
> For this particular issue, I am only providing information, and only
> because somebody else already brought it up. I find the random focus
> problem (and the possibly related random lxterminal dimensions) to be only
> a minor nuisance. What's really troubling me with 6.7.0 (I upgraded from
> 6.4.4) is that the response to the mouse is somewhat glitchy, i.e., the
> mouse pointer will often hesitate before moving. The severity is such
> that it interferes with selecting text and makes resizing windows
> challenging, and to the degree that playing a game is almost impossible.
I noticed that in KVM, mouse response is much better when using -vga
vmware rather than -vga std or cirrus. Maybe, some xorg modules switch
to "software pointer" on their own?
> I tried renicing X, and found that at -15 the problem diminishes to the
> point of just feeling a bit "off", but it is still not as smooth as I
> would expect on a computer manufactured in the last 15 years. I also
> tried other mice, and even bought a brand new high-end mouse, but that had
> no effect. This is not a problem I experienced on any machine with any
> prior Knoppix version. If I cannot fix this, I am seriously considering
> reverting to 6.4.4. :(
There is a new xorg core server (1.11) in xorg in Debian/unstable now,
which is present in Knoppix 6.7.1. Maybe this fixes something. I did not
notice a difference on my testing machines, though (which did not show a
pointer delay problem before or after, anyways).
> >> Now, if I run lxterminal from either the command prompt, or from a
> >> script,
> >> using "lxterminal --geometry=80x24", it opens mostly 79 column x 23 line
> >> terminals, with a few 81x23. (And, very rarely, something else.) In
> >> hopes of isolating the cause, I tried this with rxvt, and always got
> >> 80x24
> >> (and, unlike lxterminal, rxvt also allows position parameters, e.g.,
> >> 80x24-0+0 for the top-right corner).
> > The compiz session manager may try to remember the geometry from the
> > last time when the window was opened, with all gtk applications. It may
> > depend on if the window gets the focus or not, which is again related to
> > the "focus stealing prevention", and may be the cause why I con't
> > observe this behaviour when testing inside kvm on my slower machine.
> >> By the way, I notice that 6.7.0 knoppix-halt does not unload any
> >> modules.
> >> That eliminates the wol problem! Thanks!
> > That's the reason why I skipped module unloading in
> > /etc/init.d/knoppix-halt. Unloading modules seems to do more than
> > desired with some modules, while most driver nowadays seem to handle
> > shutdown or suspend correctly, so it may be the better choice to never
> > unload them before shutdown.
> Thanks again!
Version 6.7.1 should be up soon (you may already get it via bittorrent
from torrent.unix-ag.uni-kl.de, just checking... not yet). Please let me
know it it got better, or worse, concerning the pointer problem.