On Thu, Mar 02, 2017 at 08:44:10AM +0100, Jiri Olsa wrote:
> On Wed, Mar 01, 2017 at 06:36:39PM -0300, Arnaldo Carvalho de Melo wrote:
> 
> SNIP
> 
> >   [acme@jouet linux]$ perf stat usleep 1
> > 
> >    Performance counter stats for 'usleep 1':
> > 
> >           2.022049      task-clock:u (msec)       #    0.181 CPUs utilized  
> >         
> >                  0      context-switches:u        #    0.000 K/sec          
> >         
> >                  0      cpu-migrations:u          #    0.000 K/sec          
> >         
> >                 48      page-faults:u             #    0.024 M/sec          
> >         
> >          2,177,301      cycles:u                  #    1.077 GHz            
> >         
> >            245,106      instructions:u            #    0.11  insn per cycle 
> >         
> >             48,000      branches:u                #   23.738 M/sec          
> >         
> >              3,936      branch-misses:u           #    8.20% of all 
> > branches        
> > 
> >        0.011181563 seconds time elapsed
> > 
> >   [acme@jouet linux]$
> > 
> > ---------------------------------------------
> > 
> > Seems related to the auto :u done when the user can't do kernel profiling, 
> > etc.
> > 
> > Re-applying your patch, to triple check:
> > 
> > [acme@jouet linux]$ perf stat usleep 1 2>&1 | head -2
> > Error:
> > You may not have permission to collect system-wide stats.
> > [acme@jouet linux]$
> > 
> > ------------------------------------------------
> > 
> > Waiting for v3 :-)
> 
> oh my... I'll check

could you please try with attched change

thanks,
jirka


---
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 1320352cda04..f53f449d864d 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -2374,7 +2374,8 @@ static void setup_system_wide(int forks)
                                return;
                }
 
-               target.system_wide = true;
+               if (evsel_list->nr_entries)
+                       target.system_wide = true;
        }
 }
 

Reply via email to