> Gesendet: Donnerstag, 18. Juni 2015 um 18:29 Uhr
> Von: "Dan Espen"
> An: michael.gros...@gmx.de
> Cc: f...@fvwm.org
> Betreff: Re: FVWM: Schedule Periodic ceases working after some days
>
> michael.gros...@gmx.de writes:
>
> > Hi!
> >
> > As I reported some years earlier, the FVWM pager crashes
> > sometimes (say, it disappears sporadically when I switch
> > desktops).
> >
> > To fix this, my FVWM config has this code since Debian Wheezy:
> >
> >> AddToFunc StartFunction
> >> + I Schedule Periodic 2000 crashGuardPager
> >>
> >> DestroyFunc crashGuardPager
> >> AddToFunc crashGuardPager
> >> + I All ("minipager") Nop
> >> + I TestRc (NoMatch) crashHandlerPager
> >>
> >> DestroyFunc crashHandlerPager
> >> AddToFunc crashHandlerPager
> >> + I SetEnv SAVED_DESKTOP_NUMBER_MINI
> >> + I paintMinipager
> >
> > Every time the pager disappears, this code
> > automatically creates a new one within 2 seconds
> > (by letting the function 'paintMinipager' doing
> > this job).
> >
> > So far, so good.
> >
> >
> >
> > Now, my Xen based virtual Debian Wheezy VNC machine with
> > FVWM 2.5.30 has an uptime of 13 days and 7 hours, and once
> > again, the pager disappeared. I waited, but it didn't reappear.
>
> The current version of Fvwm in CVS is 2.66.
The last time I tried to replace the Debian packaged FVWM version
by a downloaded one was in March 2013, when I tried to use
FVWM 2.6.5 in Debian Squeeze (an air gap machine without
Internet connection). I discarded this approach when I realized
that I would need multiple days of researching on how to integrate
the Debian menu into the FVWM menu. When I use the packaged version
of FVWM, I (since Squeeze) just have to include the three lines
of code
> Read /etc/X11/fvwm/menudefs.hook
> AddToMenu MenuFvwmRoot
> + "&D. Debian Menu" Popup /Debian
into my FVWM config to get the Debian menu. Until Lenny, I even
had to do nothing. This is the reason why I still use the packaged
versions of FVWM instead of a downloaded one.
.
.
.
>
> I don't remember the Pager issue, but I guess it could be fixed.
> Did you submit a stack dump?
The thread was in August 2010:
https://www.mail-archive.com/fvwm@fvwm.org/msg01427.html
I used Fvwm 2.5.26 at Debian Lenny back then (and still use
it in 3 different VirtualBox machines).
The bug occurs in 2.5.26 (Lenny), 2.5.30 (Squeeze) and 2.5.30 (Wheezy).
The frequency of disappearance depends on the environment: I use
FVWM on different hardware, and on different virtualization environments:
- Directly on real hardware (both as 'root' and as 'user') when working
on computers sitting in front of them
- Virtualized in VirtualBox (both as 'root' and as 'user') when working
on computers sitting in front of them
- Virtualized in Xen (both as 'root' and as 'user') when working
on virtual computers running on remote servers using VNC
- It also happened in QEMU/KVM when I used it (before dumping KVM and
migrating to VirtualBox)
In some environments it occurs more often and in others more rarely,
and it depends on the geometry I assign to the respective instance of 'pager'.
When I remember right, I was not able to cause a crash dump when I tried it.
And Thomas Adam was not able to reproduce the crashing behavior of the pager
when he examined it.
Furthermore, we had the discussion whether the pager crashes or not.
There is the chance, that it actually may not "crash" but rather think
it should "stop working". In this case, it would mistakenly properly
close its window without bothering to leave a crash dump.
So, I decided to live with it. Until Debian Squeeze, I was able to use a
Geometry that either does not crash the pager or only crashes the pager
when I pointlessly (for test purposes) switch desktops in a loop (25 times
per second; depending on the keyboard rate; the configured autorepeat option,
which can be set with "xset r rate 250 25").
Since Wheezy (working in the VNC session), the bug occurs more often now,
and so I wrote the 'Schedule Periodic' hack.
The only requirement for that workaround should be that
the 'Schedule Periodic' command does not cease its working
.
.
.
> I can set up a Schedule test, but it sounds like it's going to take
> a couple of weeks to recreate the problem.
There is no hurry. If somebody finds a bug, then a fix would be nice.
If not, I will invent something creative.
My latest idea is: I switch desktops countless times a day. I could
let the desktop switching method start a method that checks whether
the 'Schedule Per