Mark Martin wrote:
> Mike Sullivan wrote:
>> Andre Molyneux wrote:
>>> Mike Sullivan wrote:
>>>>
>>>> you need two reviewers now? usually that's reserved for release builds.
>>>
>>> I thought I did. If not, I'm happy to proceed with one.
>>
>> usually the more eyes the better, so it's certainly not a problem if
>> you get more than one. but only one is generally required.
>> as we get closer to milestone builds sometimes more than one reviewer,
>> or particular reviewers, or even particular rti advocates are required -
>> but we aren't doing that at the moment unless I've missed something.
> Out of curiosity, what is the experience behind this? It doesn't seem
> intuitive to me that code going into a "continuously deliverable" code
> base would require 2 reviewers at release time and only 1 any other
> time. I'm sincerely interested in the rationale behind that. Isn't it
> the same integrated code being delivered at release time, regardless of
> when it actually integrated?
Because there's less and less time to fix any problems the closer
we get to shipping something, so a bit more scrutiny is a useful thing
in order to catch things before integration. Coupled with the likely
pressure to quickly fix the bugs themselves late in the release, this
may help compensate for any rushed work. It may not of course, but
that's been what we've done, though most of that I've seen in ON where
bug fixes are generally much more substantial and risky than in sfw
(where it's usually just build/packaging being reviewed, and any code
changes are mostly from upstream where they've hopefully been reviewed
already).
In any event it always made me feel better, so it at least has
psychological benefit :)
Mike
