Re: [BUG] perf script segfault

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

Em Tue, Mar 31, 2015 at 04:25:57PM +0200, Jiri Olsa escreveu:
> On Tue, Mar 31, 2015 at 11:13:26AM -0300, Arnaldo Carvalho de Melo wrote:
> 
> SNIP
> 
> > And setting tool->ordered_events to false;
> > 
> > What fails is perf_evlist__sample_id_all(session->evlist), I went as far as
> > looking at the perf_evlist__sample_id_all call that will find a first evsel,
> > with all its fields zeroed, i.e. at some point it finds out it is a pipe (perf
> > script < perf.data) and seems to not read the perf_event_attr attributes, etc,
> > but I need to dig deeper here to figure out this and why is it that in that
> > case we end with two 'perf script' processes when I think it should be just one, etc.
> > 
> > So, to keep the assumptions in place I'll keep my patch, the one that David
> > tested, and redo yours on top, looking as well at ways to use what is in
> > perf_session, should provide a patch soon.
> 
> I was just replying to the poluted ordered_events .. I haven't tracked

:-) Yeah, that was kinda a layering violation, its good that we can
remove it from there, as for evlist... I did a quick check and yeah, we
can remove it from ordered_events and move the  perf_evlist__parse to
the deliver event function, where it will be obtained from perf_session
that was obtained from container_of (oe), its just one statistic that
remains, but yeah, probably with some work that can as well be removed.

We could then think of this as needed while isolating those
ordered_events code out of perf_session, to make it useable by other
tools (my initial intent as for trace to use it), but in the end result,
after the deliver_event was introduced, it is not needed anymore.

Yeah, machines can as well be obtained from perf_session.

> the segfault totaly.. we can change that after your fix of course ;-)

:-)

- Arnaldo