[vtk-developers] Valgrind defects are zero!

[Bill Lorensen]

David,

I definitely agree that we should require all updates to VTK to pass
through gerrit. The peer review process helps produce better code. As
you know we already do that for ITK. Also, ITK has 3 robot builds that
are kicked off for each gerrit patch. This certainly helps but does
not eliminate the introduction of defects. We support too many
platforms and options to be able to cover all possibilities.

Bill

On Thu, Dec 8, 2011 at 2:18 PM, David Doria <daviddoria at gmail.com> wrote:
> On Thu, Dec 8, 2011 at 2:06 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>> David,
>>
>> I think the lack of support by the community is disappointing. Peer
>> pressure is one of the few effective tools that we have.
>>
>> The next step is to revive my stick. It's been in storage for a while.
>>
>> Bill
>
> I think the new Gerrit procedures could totally eliminate this type of
> problem. If the dashboard is green and there are 0 valgrind defects,
> then if we simply don't accept any code that introduces dashboard
> errors/defects, both of these quantities should stay at zero. Could
> Gerrit treat that as a requirement, so that new code is only allowed
> to merge into master if it introduces no problems? Allowing a human to
> make this decision allows laziness/oversight/forgetfulness to play a
> role.
>
> Maybe something like this:
> Automatically add a robotic reviewer named "Dashboard" to each topic
> that gets pushed to Gerrit. The current requirement to merge (a human
> saying "looks good, approved") should remain, but additionally this
> new robotic reviewer must give its approval as well. This approval
> would be "approved" for 0 errors/warnings/defects, and "I'd rather you
> not merge" otherwise.
>
> Of course this would require some work with the Gerrit team, but it
> seems like a feature that everyone should be interested in.
>
> Thoughts?
>
> David



-- 
Unpaid intern in BillsBasement at noware dot com