Re: [PATCH 3/5] perf tools: Copy metric events properly when expand cgroups

From: Namhyung Kim
Date: Fri Sep 25 2020 - 09:45:09 EST


On Fri, Sep 25, 2020 at 10:26 PM Jiri Olsa <jolsa@xxxxxxxxxx> wrote:
>
> On Thu, Sep 24, 2020 at 09:44:53PM +0900, Namhyung Kim wrote:
>
> SNIP
>
> >
> > + if (metric_events) {
> > + orig_metric_events = *metric_events;
> > + rblist__init(metric_events);
> > + } else {
> > + rblist__init(&orig_metric_events);
> > + }
> > +
> > for (;;) {
> > p = strchr(str, ',');
> > e = p ? p : eos;
> > @@ -255,6 +267,14 @@ int evlist__expand_cgroup(struct evlist *evlist, const char *str)
> > cgroup__put(cgrp);
> > nr_cgroups++;
> >
> > + if (metric_events) {
> > + perf_stat__collect_metric_expr(tmp_list);
> > + if (metricgroup__copy_metric_events(tmp_list, cgrp,
> > + metric_events,
> > + &orig_metric_events) < 0)
> > + break;
> > + }
>
> looks good, do you plan to actualy add support for record?

No actually, I still think perf record should use --all-cgroups.

> my ack from last version stays

Thanks! But I didn't see your ack for this patch set.
(I've only seen it for the perf inject patchset..)

Thanks
Namhyung