On 8/16/21 1:36 PM, Dan Carpenter wrote:
There should be a Fixes-from: tag for bugs found in review (not style
issues) but when I suggest it then people just say to use the
Reported-by tag.
I think things caught during review aren't normally worthy
of specific mention in the commit message (though maybe in
the non-committed part under "---"). I mean, that's what
review is for. And in the case of what<lkp@xxxxxxxxx>
does, that's effectively a technical aspect of "review."
I'm not talking about stuff like intending or naming schemes, I'm
talking about real bugs like missing error codes or NULL dereferences.
People do count tags so we might as well add them for worthwhile
behavior.
So you're saying that things caught during review *should* be
given credit, as opposed to acknowledging the credit for catching
it only when the bug slips by the reviewers, caught after commit.
I understand that, and I get your point about the incentives
(which take the form of tags with acknowledgement).
As I indicated earlier, I'm all for showering credit on everyone
that helps. But I still think doing so for input taken during
the review phase is too much, and full of fuzzy cases (how do you
judge whether a suggestion is worth acknowledging?).
I think what you do with Smatch is outstanding, and you deserve
a lot of credit for it. But like checkpatch.pl, it would be even
better if people used it to catch things *before* they ever went
out for review. That option would give *no* credit to Smatch for
catching problems early. Yet catching issues as early as possible
is a good thing. Should we acknowledge checkpatch.pl when it
tells us to fix something it finds; if so, which of them?
So I don't think "Fixes-from" (whatever that means) or
"Reported-by" make sense for this type of update.
Earlier today I forwarded a kbuild Smatch warning where someone had
used "sizeof(0)" instead of "0" but because the patch was already
applied, that means I got Reported-by credit. If the kbuild-bot could
have reported the bug before the networking people applied it that's
more valuable but I get less credit. It's a perverse incentive.
It's a perverse incentive for you as Smatch developer. But I think
the better place to put an incentive is on getting people to avoid
sending patches at all until they have used tools available to
automatically find issues before they get out for review.
Also I sort of don't like the Reviewed-by tag. I see a lot of people
adding Reviewed-by but I've never seen them point out a bug during the
review process so that seems pretty worthless. But Fixes-from means
that person knows what they're talking about.
That's not a problem with Reviewed-by, it's a problem with people
misusing it. Are you suggesting that "Fixes-from" would be applied
by the developer, not reviewer? Regardless, Reviewed-by is *supposed*
to carry meaning. "Documentation/process/submitting-patches.rst" has
a section that describes what the "Reviewer's statement of oversight"
represents.
I think it would be nice to recognize review feedback. It's
actually more valuable than the summary statement "I have
reviewed this and find it acceptable." But I don't believe
adding new acknowledgement tags is a good way to do it.
-Alex
regards,
dan carpenter