Re: [PATCH] perf stat: Ignore error thread when enabling system-wide --per-thread

[Arnaldo Carvalho de Melo]

Em Tue, Feb 27, 2018 at 08:53:13AM +0800, Jin, Yao escreveu:
> 
> 
> On 1/23/2018 10:19 PM, Jiri Olsa wrote:
> > On Mon, Jan 22, 2018 at 01:10:31PM +0800, Jin, Yao wrote:
> > > 
> > > 
> > > On 1/16/2018 9:17 PM, Jiri Olsa wrote:
> > > > On Tue, Jan 16, 2018 at 09:06:09PM +0800, Jin, Yao wrote:
> > > > > Just tested. But looks it's not OK for '--per-thread' case.
> > > > 
> > > > yea, I haven't tested much.. might need soem tweaking,
> > > > but my point was that it could be doable on one place
> > > > instead of introducing another if possible
> > > > 
> > > > jirka
> > > > 
> > > 
> > > Hi Jiri,
> > > 
> > > I ever considered to move the operation of removing error thread to
> > > perf_evsel__fallback(). The perf_evsel__fallback() is common code and it's
> > > shared by perf report, perf stat and perf top.
> > > 
> > > While finally I think it'd better let the caller decide to remove error
> > > thread and try again, or just return the warning message.
> > > perf_evsel__fallback() probably doesn't know what the caller want to do.
> > > 
> > > That's my current thinking. Maybe there will be a better fix...
> > 
> > ok, can't think of better fix atm.. looks good ;-)
> > 
> > Acked-by: Jiri Olsa <jolsa@xxxxxxxxxx>
> > 
> > thanks,
> > jirka
> > 
> 
> Hi Arnaldo,
> 
> Could this fix be accepted?

yeah, its just strange that that err_thread is not used, applied.

- Arnaldo