Hi,

On Tue, Apr 10, 2007 at 02:32:23PM +0200, Martin Husemann wrote:
> On Tue, Apr 10, 2007 at 02:22:14PM +0200, Gert Doering wrote:
> > I wouldn't mind if the kernel would actually just SIGSEGV/SIGBUS the 
> > offending process that produced the alignment error (and have it core 
> > dump, which is something I'm more familiar in debugging...)
> 
> It would, if the application wouldn't catch it. Find the signal/sigaction
> call that masks it, comment that out and recompile the app.

OK, will do (it's either milter-greylist, libmilter, or pkgsrc libpthreads,
"can't be more than a long weekend"... :) ).

Just to help me understand this - what is "MDP_FIXALIGN"?

From reading the code path in arch/sparc64/sparc64/trap.c, it seems
that this flag would, if set, cause the kernel to "fixalign()" the 
offending access, and not cause a SIGBUS to be raised.

But maybe I am just misunderstanding all this...

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@greenie.muc.de
fax: +49-89-35655025                        gert@net.informatik.tu-muenchen.de