Re: Motivational statistics

[David Kastrup]

Han-Wen Nienhuys <address@hidden> writes:

> On Sat, Feb 1, 2020 at 12:59 AM Urs Liska <address@hidden> wrote:
>>
>> Hi all,
>>
>> I wanted to get a better understanding from my impression of the
>> significant increase in traffic on lilypond-devel.
>>
>> For this I did some statistics on James' "PATCHES - Countdown"
>> messages. Since patches are counted multiple times while flowing
>> through the process I think the only relevant metric is the "New"
>> section, and this should not be calculated by the countdown message but
>> averaged by day. And the results are convincing:
>>
>> Four weeks leading to the Salzburg conference:
>> 0,32 new patches per day
>>
>> Since Salzburg:
>> 1,46 patches per day
>
> Thanks for the stats. I guess it's mostly me, though.
>
> $ git log --author=hanwen --since 2016 | grep ^commit| wc
>      16      32     768
>
>> Of course these are no scientifically hardened results - but they match
>> the feeling of excited frenzy visible on this list. However sustainable
>> the effect may be, the short term impact of the developer meeting and
>> the conference seems to have been remarkable.
>
> I am also forming more coherent ideas about the development process,
> but I am still unsure about the final push process. As I understand
> it, you have to push to staging, and then someone (David?) runs patchy
> over the staging branch, verifies the regtest output, and pushes to
> master. Is that roughly correct?

No.  staging moves without manual verification.  There are different
Patchy processes for staging and for issue review.  The one for staging
only checks that make, make test, and make doc all complete
successfully.  Several people run that as needed (we used to have a
computer administered by James running it regularly every two hours, but
his company rules have stopped this from being possible).

The review patchy in contrast requires visual inspection of regtest
results.  The automation of the process deteriorated significantly after
we had to stop using Google Code because the scripts have not been
adapted to the current situation, and at the current point of time it is
just James who does those tests with considerably more manual effort
than previously.

-- 
David Kastrup