Date:   Thu, 3 Mar 2022 14:27:42 +0100
From:   Johan Hovold <johan@...nel.org>
To:     Dan Carpenter <dan.carpenter@...cle.com>
Cc:     Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
        Matthew Wilcox <willy@...radead.org>, kbuild-all@...ts.01.org,
        Jonathan Corbet <corbet@....net>, linux-doc@...r.kernel.org,
        linux-kernel@...r.kernel.org, Florian Eckert <fe@....tdt.de>
Subject: Re: [PATCH v1 1/1] docs: process: submitting-patches: Clarify the
 Reported-by usage

On Thu, Mar 03, 2022 at 12:54:32PM +0300, Dan Carpenter wrote:
> On Tue, Feb 01, 2022 at 09:51:33AM +0100, Johan Hovold wrote:
> > On Mon, Jan 31, 2022 at 08:16:42PM +0200, Andy Shevchenko wrote:
> > > On Mon, Jan 31, 2022 at 05:47:32PM +0100, Johan Hovold wrote:
> > > > On Mon, Jan 31, 2022 at 05:34:35PM +0200, Andy Shevchenko wrote:
> > > > > On Mon, Jan 31, 2022 at 04:18:30PM +0100, Johan Hovold wrote:
> > > > > > On Fri, Jan 28, 2022 at 01:44:20PM +0000, Matthew Wilcox wrote:
> > > > 
> > > > > > > I think this misunderstands the problem that Andy is trying to fix.
> > > > > > > 
> > > > > > > The situation: I write a patch.  I post it for review.  A bot does
> > > > > > > something and finds a bug (could be compile-error, could be boot
> > > > > > > problem).  That bot sends a bug report with a suggestion to add
> > > > > > > Reported-by:.  That suggestion is inappropriate because the bug never
> > > > > > > made it upstream, so it looks like the bot reported the "problem"
> > > > > > > that the patch "fixes".
> > > > > > > 
> > > > > > > It's not unique to "new feature" patches.  If I'm fixing a bug and
> > > > > > > my fix also contains a bug spotted by a bot, adding Reported-by
> > > > > > > makes it look like the bot spotted the original bug, rather than
> > > > > > > spotting a bug in the fix.
> > > > > > > 
> > > > > > > The best thing to do in this case is nothing.  Do not credit the bot.
> > > > > > > Maybe add a Checked-by:, but that would be a new trailer and I really
> > > > > > > don't think we need a new kind of trailer to get wrong.
> > > > > > 
> > > > > > It seems like the only way to fix this is to fix the bots. Adding more
> > > > > > documentation is unlikely to help in this case.

> > Perhaps I'm missing something, but if you re-read Mathews description
> > above, it still seems to me like the issue is that the bots are trying
> > to claim credit for finding things that haven't been merged yet.
> > 
> > Your suggestion is to document that the bots should be ignored. My
> > suggestion is to fix the bots.
> 
> Originally the kbuild bot used to not have that notice but adding it
> meant that kbuild bot got a lot more visibility.  The truth is that
> managers love metrics and it helps people get paid.
> 
> The whole point of kbuild-bot was to search the lists and test code
> before it gets merged.  If they just waited and tested linux-next they
> would get their reported by tags because most trees don't rebase.  But
> we're punishing them for being better at their job.  It's a perverse
> incentive.

I hear you. But I also get Matthew's and Andy's point about it not being
quite right to give an automatic test program Reported-by credit for
finding, say, an unused variable in a not yet merged patch. And perhaps
even more so since real reviewers often get no credit at all (but
perhaps a mention in a cover letter).

> We should create a new tag for finding bugs during review.

Yeah, I guess your perverse incentive argument applies equally to human
reviewers even if I'm also reluctant to going down this path.

Johan

Powered by blists - more mailing lists