This is the mail archive of the
libc-ports@sources.redhat.com
mailing list for the libc-ports project.
Re: setting of initial bottom of stack variable in pthread.c (again)
- From: Jay <j2list468 at yahoo dot com>
- To: libc-ports at sourceware dot org
- Date: Tue, 22 Aug 2006 08:59:44 -0700 (PDT)
- Subject: Re: setting of initial bottom of stack variable in pthread.c (again)
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=1S/R77DveHWsQanMFjpdXF8PfedTTd6JddeF6KXbCB6l4eCVkXS153/e3W/RhTtfHaHAFZ8UBlSMJbooiZq7uDGhhiYMz0kG5zqUd855OBAXCTcG7bV9asipGifus+tUQh1F+sPhz9WtibzUKMNJCHpSVeZRiTzUffsW8TiNdsA= ;
I did a grep and found that THREAD_STACK_START_ADDRESS
was defined in
glibc/linuxthreads/sysdeps/arm/pt-machine.h to
0xC0000000 - STACK_SIZE which in our case is
0xBFE00000. I downloaded glibc2.4 & 2.3.6, but it
looks as if pthreads has been removed in them. The
linux distribution we are using uses an older version
of glibc and it looks as if they modify things every
once in a while. I started looking in the thread
library since they did modify something there, but it
wasn't anything that would cause this problem. In any
case, I did comment out the defining of
THREAD_STACK_START_ADDRESS (went back to an older
version of their distribution and found defining
THREAD_STACK_START_ADDRESS was not in pt-machine.h in
the older version), and this took care of
THREAD_STACK_START_ADDRESS being different than
__pthread_initial_thread_bos.
Thanks for all your help.
Jay
--- Daniel Jacobowitz <drow@false.org> wrote:
> On Tue, Aug 22, 2006 at 08:22:50AM -0700, Jay wrote:
> > I put some printf's into the thread segment
> function.
> > For some reason THREAD_STACK_START_ADDRESS =
> BFE00000
> > while __pthread_initial_thread_bos = BFA00000, so
> I
> > believe I need to figure out why these two are
> > different. My first guess is going to be that
> > THREAD_STACK_START_ADDRESS is defined somewhere
> else
> > so it never gets set to
> __pthread_initial_thread_bos
> > as in internals.h
>
> Try compiling manager.c, with the same command line
> as the glibc
> build log, but adding -dD -E in place of -c. The
> resulting text file
> should show you where it's being defined. In an
> unmodified FSF source
> tree, I can't see anywhere this could happen!
>
> --
> Daniel Jacobowitz
> CodeSourcery
>
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com