Re: [PATCH 3/9] perf hists: Add support for showing relative percentage

[Arnaldo Carvalho de Melo <acme@ghostprotocols.net>]

Em Wed, Mar 19, 2014 at 12:18:00AM +0000, Namhyung Kim escreveu:
> Hi Arnaldo,
> 
> On Tue, Mar 18, 2014 at 8:08 PM, Arnaldo Carvalho de Melo
> <acme@ghostprotocols.net> wrote:
> > Em Mon, Mar 10, 2014 at 04:43:54PM +0900, Namhyung Kim escreveu:
> >> @@ -695,12 +695,17 @@ static void hists__remove_entry_filter(struct hists *hists, struct hist_entry *h
> >>       if (h->filtered)
> >>               return;
> >>
> >> -     ++hists->nr_entries;
> >> -     if (h->ms.unfolded)
> >> +     hists->nr_entries++;
> >> +     hists->nr_non_filtered_entries++;
> >
> > Why not keep existing practice? prefix or suffix generates the same
> > code, changing it from prefix to suffix increment just adds noise to the
> > patch :-\
> 
> Argh, sorry - I couldn't resist changing it.  Won't do that in the future. :-/

We should strive to be as minimalistic as possible. Reviewing is hard,
we shouldn't make it harder :-)

Cleanup/cosmetic patches, when deemed necessary should be done
separatetly and stated as such.
 
> > Also, this is why I was asking about nr_entries and total_period being
> > invariant, looking at this function we can see it is _not_ invariant, as
> > when we apply/remove filters we touch them.
> >
> > This is why I made the comment (in a private conversation) about having
> > a invariant_total_period invariant_nr_entries pair (I think I used some
> > other naming) while knowing that the existing variables nr_entries and
> > total_period are actually subjected to the filters being used.
> >
> > I.e. to avoid confusion we need to make total_entries and nr_entries
> > never change when a filter is applied, touching only two new variables
> > for non_filtered total_period and nr_entries.
> >
> > I'll try doing it if you don't do it first, but will first process some
> > more patches and submit what I already processed.
> 
> Please see the next patch 4/9 ("perf report: Add --percentage option")
> doing that.  What I did in this patch is just adding new
> non_filtered_* fields and not changing existing behavior.  And in the
> next patch, the total_period and nr_entries will become invariant.
> 
> Do you want to split the patch 4/9 or merge a part of it into this?

The point is that I don't look at all the patches (nor should I, I
think) to figure out what one wants to do after a series is applied.

So each patch should stand on its own, each one should be meaningful
without requiring reading what is coming next.

- Arnaldo