Re: [Xen-devel] [RFC] Results of Phase 1 of the Review Process study

[Lars Kurth <lars.kurth.xen@xxxxxxxxx>]

> On 16 Oct 2015, at 12:06, Stefano Stabellini 
> <stefano.stabellini@xxxxxxxxxxxxx> wrote:
> 
> On Wed, 14 Oct 2015, Lars Kurth wrote:
>> Hi everyone,
>> 
>> I wanted to walk you through some of the results of the Review Process study 
>> https://github.com/dicortazar/ipython-notebooks/blob/master/projects/xen-analysis/Code-Review-Metrics.ipynb
>>  
>> 
> 
> I noticed that the "time to ack" doesn't seem to be particularly
> increasing over time (Out35), while both "time to merge" (Out20) and
> "time to commit" (Out23) are increasing.  Is it a fair observation?

This looks a little bit unusual. I was wondering why time to commit was 
increasing, but originally was thinking this may have something to do with the 
release cycle, given that the median (green) has stayed steady and there were a 
few spikes in the mean (blue)

E.g.
- ACK'ed, forgotten and applied before the release
- ACK'ed before freeze, applied after
 
But of course comments and re-review after ACKs are also an explanation.

> If the observation is correct, I wonder why. Could it be because the
> committers are too busy? Or maybe could it be because the number of
> comments after a series has already been acked by somebody, but before
> it is committed, is increasing over time?
> 
> 
> Maybe in reality the workflow is:
> 
> 1---------2----------3---------4--------5---------A--------6--------7--------C
> 
> 
> Could we measure the number of iterations or the number of comments
> after a series has been acked (between A and C)?

I suppose there are two possible ways to measure this: 
- Additional revisions between A and C
- Comments sent after A : although some of them may of course be comments 
related to committing (e.g. a note that it has been applied, ...)

It may also possible that these are cases where we had test failures on the 
push gate, leading to further revisions, after an attempt to commit. But that's 
not 100% clear to me: I think we do only try to match changes going into master.

Regards
Lars
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel