Re: OpenBSD and IdentifyPhone function (fwd)

[BORBELY Zoltan]

Hi,

On Sat, Jan 10, 2004 at 07:52:44PM +0100, Pawel Kot wrote:
> > The following fix committed into the cvs, please test it:
> >
> >     
> > http://savannah.nongnu.org/cgi-bin/viewcvs/gnokii/gnokii/common/phones/nk6100.c.diff?r1=1.161&r2=1.162
> 
> Sorry, I can't see what it is supposed to fix. Segfault comes after
> PhoneInfo return. And revision is filled in with snprintf. So how can this
> change help?

This bug was quite tricky... :-) Under OpenBSD everything is compiled
with ProPolice, which is a stack protection system. The abort was caused
by ProPolice when the IdentifyPhone() function tries to return. It means
the stack frame of the IdentifyPhone() was corrupted.

Look into the the 0xc8 case of the IncomingSecurity() function. If
data->revision[0] isn't 0, the ", SW foo" or ", HW foo" string will
be contcatenated. If data->revision is initialised, everything will
be working fine.

Bye,
Bozo