Re: [PATCH RFC 1/2] MAINTAINERS: Add "R:" designated-reviewers tag

[Frank Rowand <frowand.list@gmail.com>]

On 6/4/2014 9:01 PM, Dave Chinner wrote:
> On Tue, Jun 03, 2014 at 01:43:47PM -0400, Steven Rostedt wrote:
>> On Tue, 3 Jun 2014 17:16:54 +1000
>> Dave Chinner <david@fromorbit.com> wrote:
>>
>>
>>> If you take it to an extremes. Think about what you can test in 15
>>> minutes. Or for larger patchsets, how long it takes you to read the
>>> patchset?
>>
>> Yeah, what about that?
> 
> That testing a patch for obvious, common regressions takes no longer
> than it does to read and review the logic. 
> 
>>> IMO, every reviewer has their own developement environment and they
>>> should be at least testing that the change they are reviewing
>>> doesn't cause problems in that environment, just like they do for
>>> their own code before they post it for review.
>>
>> Let me ask you this. In the scientific community, when someone posts a
>> research project and asks their peers to review their work. Are all
>> those reviewers required to test out that paper?
>> Or are they to review it, check the math, look for cases that are
>> missed, see common errors, and other checks? I'm sure some
>> reviewers may do various tests, but others will just check the
>> logic. I'm having a very hard time seeing where Reviewed-by means
>> tested-by. I see those as two completely different tags.
> 
> We are not conducting a scientific research experiment here. We are
> conduting a very large software *engineering* project here.

Yes, software engineering.  Where software review is a manual process
of *reading* and understanding code, in all of the processes I have
been involved in at big corporations that love big process.  (Not to
claim I know of all the processes everyone else uses...)

Why can't you just let reviewed-by and tested-by mean different
things instead of one being a super-set of the other?

If you force reviewed-by to also mean tested-by then you just
shrank your available pool of reviewers.

</dead-horse beating>

> 
> So perhaps we should be using robust software engineering processes
> rather than academic peer review as the model for our code review
> process?

< snip >

Cheers,

Frank