Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email firstname.lastname@example.org
From: Michal Jaegermann (michal_at_ellpspace.math.ualberta.ca)
Date: Mon Oct 14 2002 - 20:38:03 CDT
On Mon, Oct 14, 2002 at 05:50:18PM -0700, Ram Viswanadha wrote:
> I upgraded to RH 7.2 (stock kernel and GCC) and now I am not seeing any
> unaligned trap messages.
> I read Rich Payne's mails in the archives and they suggest that the problems
> are in users code not in the kernel.
> So there are two possibilities, since my code has not changed:
> a) The kernel has become more lenient and is not producing any unaligned
> trap messages.
> b) GCC had a bug which produced unaligned data which has hence been fixed.
Most likely seems to be the third option - printk in a kernel code which
was producing these messages was turned off. :-)
Unless you can show otherwise assume that "unaligned access" means that
your code has bugs. Even if your kernel is silent about these, which
does not mean at all that it is more "lenient", they are still bugs.
On many ocasions "unaligned access" messages are side-effects, not
causes, of killer bugs. Other times your kernel will trap, do a fixup,
and move along but you will get a performance hit. If this happens
once per some infrequently used program this is not a big deal.
If this is in some widely used library then this may be significant.
Axp-list mailing list