[vtk-developers] Requiring an experimental dashboard

[Bill Lorensen]

Yes, as you know we do this for itk. Automatic, if the developer has
approved credentials. I assume the same will be done for vtk sometime.

On Wed, Dec 28, 2011 at 1:47 PM, David Doria <daviddoria at gmail.com> wrote:
> Could we require (or at least suggest) an experimental dashboard is run
> before a Gerrit patch is merged, or perhaps even before a branch can be
> pushed to Gerrit? Maybe this could be built into the gerrit-push and/or
> gerrit-merge scripts?
>
> Here is the current workflow:
>
> 1) A developer writes some code and pushes it to Gerrit. This is allowed
> whether or not the code compiles, runs, or passes all tests.
> 2) The developer potentially either a) forgets to run tests at all or b)
> only runs tests that he thinks are relevant to the modifications
> 3) Reviewers either a) assume the developer ran the right tests so they
> don't run their own or b) reviewers forget to run tests or c) reviewers also
> run the wrong subset of tests
>
> The workflow I would suggest is:
>
> 1) A developer writes some code.
> 2) When trying to push to Gerrit, the developer is interrogated by a script
> that says "Did you run an experimental dashboard? Did you run all tests (if
> the dashboard was not run)? etc., etc. Ideally it would allow him to take
> action from within the script, but if nothing else at least it forces him to
> say "no I did not, but I want to proceed anyway".
> 3) The list of things that was determined to have been done is included with
> the commit message or otherwise displayed on Gerrit.
> 4) Reviewers can review the code already knowing which measures have already
> been taken to verify that the code is clean.
>
> Ideally the 'request build' idea from the ITK Gerrit would be automatically
> invoked when pushing to Gerrit, but of course it takes significant hardware
> resources to support many builds. The idea of my proposition is that it
> leaves the burden on the developers machine, but takes "I forgot" out of the
> list of excuses why something is wrong. Taking the human element out of it
> always seems like a good idea to me.
>
> David
>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtk-developers
>
>



-- 
Unpaid intern in BillsBasement at noware dot com