[vtk-developers] Valgrind defects are zero!

[Bill Lorensen]

Dave C,

I agree. There is no way we can run valgrind for each gerrit topic.
Since it is at zero now, we should be able to keep it there with
little effort. For years I used to check the vtk dashboard with my
morning coffee. For the past few years I've spent most of my coffee
time with itk.

For itk we have a shell script that a developer can run to check
valgrind on an individual test. Maybe we should provide the same for
vtk.

I'll look into it,

Bill

On Thu, Dec 8, 2011 at 3:12 PM, David Cole <david.cole at kitware.com> wrote:
> On Thu, Dec 8, 2011 at 2:33 PM, David Doria <daviddoria at gmail.com> wrote:
>>> We support too many platforms and options to be able to cover all possibilities.
>>
>> Of course we can't cover ALL possible configurations. However, the
>> only defects we know about (the ones you fixed) are the platforms we
>> test for. Therefore it seems to make sense to only allow new code that
>> does not introduce problems on platforms we support/test, right?
>>
>>> Also, ITK has 3 robot builds that are kicked off for each gerrit patch.
>>
>> Right, but the *assumption* is that the human reviewers will look at
>> this, which may or may not happen in practice. What I am suggesting is
>> that these robot builds report back to gerrit (as "robot reviewers")
>> and flag the patch as "not approved" if anything fails/has defects.
>>
>>> This certainly helps but does not eliminate the introduction of defects.
>>
>> My proposal *eliminates* the introduction of defects on platforms we
>> test, which is really the best we can hope to do.
>> _______________________________________________
>> 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
>>
>
> All fine suggestions in theory, but running a valgrind dashboard on
> every single gerrit topic will be quite time consuming. We will need
> multiple cdash at home clients running dashboards in order to keep up
> with the volume of commits if we expect a valgrind dashboard from each
> and every gerrit patch.
>
> And Bill's point remains valid: there will still be untested
> combinations (in fact, I'm sure there are untested combinations right
> now) that will still allow defect introduction.
>
> We can definitely do better, but the only way to guarantee not adding
> any defects is to stop development work. :-)



-- 
Unpaid intern in BillsBasement at noware dot com