Re: [PATCH v10 22/29] perf s390-sample-raw: Don't pass evsel or its PMU with sample

[Ian Rogers]

On Mon, Apr 13, 2026 at 1:21 PM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
>
> On Mon, Apr 13, 2026 at 09:59:56AM -0700, Ian Rogers wrote:
> > On Mon, Apr 13, 2026 at 9:41 AM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
> > >
> > > On Sat, Apr 11, 2026 at 07:08:25PM -0700, Ian Rogers wrote:
> > > > The sample contains the evsel and its PMU so it is unnecessary to pass
> > > > the evsel as well.
> > >
> > > This patch is changed to minimize the diff.
> >
> > I'm not sure what this means as the patch already attempted to
> > minimize the diff. I'm worried that not sending changes to my patches
> > via the list introduces errors because they don't go through Sashiko.
>
> Please take a look at the tmp.perf-tools-next branch.  I kept the local
> evsel variable.

Right, your change looks okay. However, based on past Sashiko reviews
(you may get similar results if you post the patches) I had been
avoiding having sample->evsel and evsel, especially when the evsel was
mutated. This happened because one would get out of sync with the
other, and then Sashiko would (rightly) warn about the divergence. You
don't diverge if there's only one. I think in this case there are
about 6 lines changing `evsel` to `sample->evsel`, while you need one
line to reassign the evsel back to `sample->evsel`—the error prone
pattern I was prompted to remove.

Thanks,
Ian

> Thanks,
> Namhyung
>