Re: [PATCH][RFC] tracing: Use class->reg() for all registering of events

[Arnaldo Carvalho de Melo <acme@infradead.org>]

Em Wed, Jun 09, 2010 at 01:53:34PM +0200, Peter Zijlstra escreveu:
> On Wed, 2010-06-09 at 17:17 +0530, Srikar Dronamraju wrote:
> > * Peter Zijlstra <peterz@infradead.org> [2010-06-09 08:32:30]:
> > 
> > > On Wed, 2010-06-09 at 10:35 +0530, Srikar Dronamraju wrote:
> > > 
> > > > Peter, can you please point me to the fix for the floating point
> > > > exception patch?
> > > 
> > > Ingo just commited the fix, see tip commit f6ab91add6 (perf: Fix signed
> > > comparison in perf_adjust_period()).
> > 
> > Actually this fix was already checked-in but I still see the problem.
> > 
> > When ran gdb on the core, it showed me that it was dumping in
> > /usr/lib64/libnewt.so.0.52. in newtScaleSet function.
> > My newt lib was of version : newt-0.52.2-15
> > 
> > I am now able to workaround the problem by uninstalling newt-devel.
> 
> Ah, ok, then we're talking about two different issues, the patch I
> referred to solves a possible /0 in the kernel which triggers easily
> with software events because of that unsigned comparison.
> 
> I'll have to defer to Arnaldo on the newt issue.

I'm trying to reproduce this now.