[Vm-dev] [unix] 4.10.2.2614 segfaults on ARM

David T. Lewis lewis at mail.msen.com
Fri Oct 11 01:35:51 UTC 2013


On Thu, Oct 10, 2013 at 05:07:11PM -0700, tim Rowledge wrote:
> 
> 
> On 10-10-2013, at 3:40 PM, Ian Piumarta <piumarta at speakeasy.net> wrote:
> > On RPi the drgeo.image runs fine with both the latest RPi binary (accelerated bitblt) and a VM compiled from the source archive with the same version number that Bert identified in the subject line.  I tested this with both local (via tightvncserver) and remote displays.
> > 
> > piumarta at raspberrypi:~/Squeak-4.10.2.2614-src$ bld/squeakvm -version
> > 4.10.2-2614 #1 Thu Oct 10 21:51:05 UTC 2013 gcc 4.6.3
> > Linux raspberrypi 3.2.27+ #250 PREEMPT Thu Oct 18 19:03:02 BST 2012 armv6l GNU/Linux
> > plugin path: /home/piumarta/Squeak-4.10.2.2614-src/bld/ [default: /home/piumarta/Squeak-4.10.2.2614-src/bld/]
> 
> 
> Well that's interesting; on my Pi it segfaults.
> 
> My version info (paraphrased)
> 4.10.2-2781
> raspberrypi 3.6.11+ #456 PREEMPT
> 
> So something between those two would appear to trigger a problem.
>

The SVN version difference (2781 versus 2614) might be a red herring. I
think that Ian built a 4.10.2-2614 level VM to match that of the VM from
the original problem report on OLPC. It works when Ian runs DrGeo on that
level of VM, but not on the OLPC. So there may be something else going on
here unrelated to the SVN version level.

It may turn out to be related to the address at which the operating system
allocates the object memory, which may vary for different OS flavors and
hardware configurations. As Ian pointed out: "It looks like the heap was
allocated in the upper half of memory, which could be a clue. :)"

Dave
 


More information about the Vm-dev mailing list