Message-ID: <email address hidden>
Date: Sun, 25 Jan 2004 12:26:25 +0100
From: Christian Kurz <email address hidden>
To: Paul Slootman <email address hidden>, <email address hidden>
Cc: Javier =?iso-8859-1?Q?Fern=E1ndez-Sanguino_Pe=F1a?= <email address hidden>
Subject: Re: Bug#220928: john: Does not build in Sparc
On [18/12/03 17:35], Paul Slootman wrote:
> On Sun 16 Nov 2003, Christian Kurz wrote:
> > [...]
> > > Benchmarking: Standard DES [24/32 128K]... make[3]: *** [sparc.h] Err=
or 1
> > > So it seems to break in the 'bench' program (in some of the '||exit 1=
')=20
> > > statements there. There might be two ways to fix this:
> It segfaults when setting the SIGALRM signal handler at line 100 of
> bench.c. This indicates to me that some sort of memory corruption has
> gone on before that.
Oh, that's not nice for me to read, considering the important task that
john does.
> > investigate the build who has some knowledge about sparc. Since the
> > build logs show that 1.6.17 and earlier versions did build fine on
> > sparc. So I suppose it's some change in the build daemon or on sparc
> > that is responsible for this break.
> With memory corruption, it could be all sorts of things that trigger it.
> Assumedly the overwriting of some buffer happens all the time, but only
> now does it touch an area that's critical. This could be due to other
> libc, other compiler, other kernel...
> I'd recommend running on some sort of malloc debugger such as electric
> fence or so. Doing this on i386 should also be effective.
Okay, we'll try it and see if that helps. I'll maybe try valgrind later.
Message-ID: <email address hidden> 1?Q?Fern= E1ndez- Sanguino_ Pe=F1a? = <email address hidden>
Date: Sun, 25 Jan 2004 12:26:25 +0100
From: Christian Kurz <email address hidden>
To: Paul Slootman <email address hidden>, <email address hidden>
Cc: Javier =?iso-8859-
Subject: Re: Bug#220928: john: Does not build in Sparc
--D9sZ58tf58331Q5M Disposition: inline Transfer- Encoding: quoted-printable
Content-Type: text/plain; charset=us-ascii
Content-
Content-
[Cc'ing Javier, the co-maintainer.]
On [18/12/03 17:35], Paul Slootman wrote:
> On Sun 16 Nov 2003, Christian Kurz wrote:
> > [...]
> > > Benchmarking: Standard DES [24/32 128K]... make[3]: *** [sparc.h] Err=
or 1
> > > So it seems to break in the 'bench' program (in some of the '||exit 1=
')=20
> > > statements there. There might be two ways to fix this:
> It segfaults when setting the SIGALRM signal handler at line 100 of
> bench.c. This indicates to me that some sort of memory corruption has
> gone on before that.
Oh, that's not nice for me to read, considering the important task that
john does.
> > investigate the build who has some knowledge about sparc. Since the
> > build logs show that 1.6.17 and earlier versions did build fine on
> > sparc. So I suppose it's some change in the build daemon or on sparc
> > that is responsible for this break.
> With memory corruption, it could be all sorts of things that trigger it.
> Assumedly the overwriting of some buffer happens all the time, but only
> now does it touch an area that's critical. This could be due to other
> libc, other compiler, other kernel...
> I'd recommend running on some sort of malloc debugger such as electric
> fence or so. Doing this on i386 should also be effective.
Okay, we'll try it and see if that helps. I'll maybe try valgrind later.
Christian www.debian. org)
--=20
Debian Developer (http://
1024D/B7CEC7E8 44BD 1F9E A997 3BE2 A44F 96A4 1C98 EEF3 B7CE C7E8
--D9sZ58tf58331Q5M pgp-signature Disposition: inline
Content-Type: application/
Content-
-----BEGIN PGP SIGNATURE-----
Tp+EACgkQHJju87 fOx+glcgCdHgRkG PCQwtHCEVKJPA+ mZ6/m vP5qbHChpvS/ +kQjz
Version: GnuPG v1.3.4-cvs (GNU/Linux)
iEYEARECAAYFAkA
Dv8AnjyPrwtQiHX
=rIeu
-----END PGP SIGNATURE-----
--D9sZ58tf58331 Q5M--