Bug#423425: [malattia@linux.it: Re: xserver-xorg-input-evtouch: Changes to 'refs/tags/0.8.5']
----- Forwarded message from Mattia Dongili <malattia@linux.it> -----
Subject: Re: xserver-xorg-input-evtouch: Changes to 'refs/tags/0.8.5'
From: Mattia Dongili <malattia@linux.it>
Date: Wed, 13 Jun 2007 18:56:54 +0900
To: Jeremy Powell <jeremy@thorcom.co.uk>
Message-ID: <20070613095654.GA15475@inferi.kami.home>
X-Operating-System: Linux 2.6.22-rc4-mm2-1 i686
User-Agent: Mutt/1.5.13 (2006-08-11)
X-Disclaimer: Buh!
Hello!
On Wed, Jun 13, 2007 at 10:08:49AM +0100, Jeremy Powell wrote:
> Hi,
>
> I'm trying to fix a bug in evtouch where by if the screen is tapped and
> the date is set (backwards) at the same time then X.org crashes (which
> is more often than you might think if you want to run ntp). This is not
Woah!! it looks like it's related to this bug:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=423425
I'm bitten by this as well, pretty seldom for me actually and I noticed
it corresponding with high disk IO, but couldn't actually find the exact
cause.
> reproducable with a regular mouse. Not being that experienced with
> development of this kind of thing I can't track down the exact point in
> the code where it is occuring. gdb didnt return any useful info for me,
> maybe because its a module?
no, most probably because it's not compiled with debug informations.
Are you using debian? if so I can provide a package with all the
debugging symbols and stuff, otherwise you may want to simply build the
driver with
make CFLAGS="-g -DEVDBG"
so if you're also looking for more output you can run X with
startx -- -logverbose 8
to also better understand the FSM state transition in libtouch.c
Just copy the resulting evtouch.so into the xorg input modules dirs.
BTW: when tapping the most of the time the FSM deals with *maybetap*
states
> I have emailed the author and had little response (he reports he is very
> busy at the moment) and so am hoping you may be able to help. I have
> tried wrapping the GetTimeInMillis function to make it monotonic
> regardless of clock changes - this had no effect so I think the problem
> lies with the timers.
hmmm, so a race condition basically... this may well be in the x server
as well.
> Anyway, if you have the time to take a look, the problem is best
> reproduced by tapping the screen and setting the date via the console
> simultaneously (pressing enter to change the date ever so slightly after
> tapping the screen). With some experimenting you should be able to
> reproduce the crash nearly every time.
will surely try, I need the latop alive currently, but later I may have
some more time ;)
BTW: if you think the I pointed out is related, do you mind if I forward
the discussion to the bug tracking system? it's easier this way to ask
for some insight to xorg developers as well.
Thanks
--
mattia
:wq!
----- End forwarded message -----
--
mattia
:wq!
Reply to: