Re: [Ksummit-discuss] Nominating Fengguang Wu - 0-day

[Fengguang Wu <fengguang.wu@intel.com>]

Hi Jiri,

On Fri, Jul 29, 2016 at 01:38:36AM +0200, Jiri Kosina wrote:
>On Fri, 29 Jul 2016, Fengguang Wu wrote:
>
>> I need your understand that there are severe limitations what useful
>> contents we can provide due to the nature of merge-test-bisect. For
>> example, there will be no build logs specific for your branches/commits
>> because the extensive build tests are possible only because your code
>> are tested together with a large number of random others.
>
>I think this'd be an interesting thing to discuss per se, as my personal
>feeling is that the expectations of maintainers / git tree owners might be
>slightly diverging from reality.
>
>The reason I am bringing this up is a very recent example here:
>
>	https://lkml.org/lkml/2016/7/28/212
>
>The 0day bot identified my patch as a build-breaker, and it took me quite
>some time to figure out that this has been actually broken for years
>already (IOW yes, the treee fails to build with some configs after my
>patch has been applied, but I'm pretty sure it breaks with the same config
>even without my patch).

Yeah we may be running into an interesting case here: the build error
is true and bisect for that build error is correct. The robot end up
happily sent report for an innocent patch which happen to trigger a
more general problem.

>Hence, if you could explain in a little bit more detailed way how
>developers should interpret the reports, I'd be very grateful for such
>session.
>
>Special focus should probably be put on regressions -- once reported by
>your bot, I'd love to know the last-known-good base to compare against.

In bisect POV, the parent commit builds fine (all the way to vmlinux)
-- so it is the last-known-good base. If follow the reproduce steps in
the report email, you should be able to confirm that.

The main problem here is, first-bad-commit may not necessarily mean
root cause. It'll require human analyzes to reach such a conclusion.

>Oh, and thanks a *lot* for all the 0day efforts, I am pretty sure that it
>increases the speed of development while maintaining the quality, and a
>status update from you should IMO always be a welcome KS topic,

Thank you! :)

Regards,
Fengguang