Re: [GIT PULL] perf tools changes for v6.5

[Namhyung Kim <namhyung@kernel.org>]

Hi Linus,

On Fri, Jun 30, 2023 at 11:43 AM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> On Wed, 28 Jun 2023 at 15:00, Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > Please consider pulling perf tool changes.  I'm doing this on behalf of
> > Arnaldo as he's on vacation.  This is my first pull request and hope I
> > didn't break anything. :)
>
> Things look normal, although I find your pgp key situation confusing.
>
> I looked up your gpg key from the kernel.org pgp key repo, but that's
> apparently not the one you used for signing this.
>
> So I have two keys for you, and neither of them are then signed by
> others (although that part is probably just the usual pgp mess with
> signatures being dropped due to flooded bogus signatures, which has
> made the whole pgp infrastructure be essentially close to useless).
>
> Anyway, I wish the pgp key situation would be better, but it's not an
> actual problem.

Sorry about that.  Unfortunately I lost my old key due to a problem
in my backups.  So I had to create a new one and asked signing
by others including Arnaldo and Jiri.  Probably it's a problem in
the pgp system as you mentioned, I don't know enough about
that though.


>
> HOWEVER.
>
> What _is_ a problem is that the end result doesn't build cleanly.
>
> The
>
>         if (list_empty(&pmus))
>                 perf_pmu__scan(NULL);
>
>         ibs_pmu = perf_pmu__find("ibs_op");
>
> in the amd-ibs-via-core no longer works, and it seems that it should just be
>
>         ibs_pmu = perf_pmus__find("ibs_op");
>
> That's fine - I can do (and did) that merge resolution - and it's part
> of my normal "merge and test build", but I'm slightly unhappy that I
> wasn't told about this part.
>
> The actual data conflicts were trivial. But this was a semantic
> conflict that was invisible to git, but showed up in build testing.
>
> And I *think* you should have known about it, because the conflict you
> do talk about shows that
>
>  +perf-y += amd-ibs-via-core-pmu.o
>
> as part of the actual data conflicts.
>
> Anyway, I've resolved this, and it wasn't a *problem*, and I'm just a
> bit unhappy that it took me by surprise.

Sorry about the inconvenience.  I'll check this kind of
problem more seriously next time and let you know.

Thanks,
Namhyung