Re: remote debugging stubs (EXC_BAD_ACCESS)

[Jeroen Dekkers]

On Sat, Sep 13, 2003 at 10:46:57PM +0200, Marco Gerards wrote:
> Greg Buchholz <sleepingsquirrel@sleepingsquirrel.org> writes:
> > [Switching to thread 202.1]
> > (gdb) info threads
> > * 1 thread 202.1  0x0808d2dc in mach_msg_trap ()
> > (gdb) break diskfs_init_diskfs
> > Breakpoint 1 at 0x804b235: file init-init.c,
> > (gdb) c
> > Continuing.
> > 
> > Program received signal EXC_BAD_ACCESS, Could not access memory.
> > 0x080842ad in _hurd_exit ()
> 
> I haven't read its sourcecode, but I assume exit sends a signal to
> itself (this is how exit usually works AFAIK).  If there is no signal
> thread it can't send a signal to itself :).
> 
> So the question is: Why isn't there a signal thread?  I assume glibc
> needs to start it if you are interested in debugging this you should
> have a look there.
> 
> I have just one guess why there is no signal thread; exec wasn't
> started yet.  Perhaps this is unrelated, I hope someone with more
> knowledge about the Hurd can tell me how this works (Marcus? ;)).

Looking at the code, _hurd_exit wants to tell the proc server its exit
status, but of course the proc server isn't started yet.

Jeroen Dekkers