[vtk-developers] C++ unit tests

Tue Jul 28 14:22:18 EDT 2015

We've got years of layers of great but poorly described regression test
infrastructure in the code today that makes it harder than it could be to
make new tests. If this is yet another layer, I'm not in favor.

But if this makes adding tests tractable for newbies and we put together
and post an easy but comprehensive recipe for adding tests along with this,
then I'm very much in favor.

David E DeMarle
Kitware, Inc.
R&D Engineer
21 Corporate Drive
Clifton Park, NY 12065-8662
Phone: 518-881-4909

On Tue, Jul 28, 2015 at 2:10 PM, Bill Lorensen <bill.lorensen at gmail.com>
wrote:

> I have been adding unit tests lately using old school vtk testing.
> I'll be willing to kick the tires on David's framework.
>
> My goal in these unit tests to get as close as 100% coverage as
> possible, verifying results when possible. This includes exercising
> Error and Warning code and checking the messages themselves.
>
> I'll admit that I'm skeptical, but hopeful that adding a new test
> framework will increase code coverage.
>
> On Tue, Jul 28, 2015 at 1:34 PM, David Thompson
> <david.thompson at kitware.com> wrote:
> > Hi David(s),
> >
> > I really like the single-header idea, especially the macros for testing
> whether values match to within a tolerance.
> >
> > My concerns are
> >
> > 1. The C++ Testing/Rendering/vtkTesting.h class which provides (or now
> that we have ExternalData, provided?) access to baseline images doesn't
> seem to fit into the FIXTURE macro. Is there a way to integrate it? Or a
> way to extend the vtkTesting class to be more usable by including methods
> similar to your macros?
> >
> > 2. One thing I dislike about most/all of the C++ test frameworks is that
> the error messages are assumed to be fixed and the methods often do not
> present a way to include the incorrect result as part of the failure
> message. It is nice that your macros include that. However, given that
> tests often fail on machines I don't have access to, it is really nice to
> encourage people to write useful messages. Macros like CHECK_CLOSE would be
> better if they accepted a message as a final argument, especially if the
> argument was passed to a std::ostream object the way
> vtkGenericWarningMacro() does so that people can (are forced to) provide
> more context about where the two numbers being compared come from.
> >
> >         David
> >
> >> On Jul 28, 2015, at 10:43 AM, David Gobbi <david.gobbi at gmail.com>
> wrote:
> >>
> >> Hi David (and David),
> >>
> >> The only advantage of my unit test header is that it is a single header
> >> (a short one, too) with zero dependencies.  So it's easy to add to VTK
> >> and it's trivial to maintain.
> >>
> >> If either of you has a better solution, then go ahead and make a merge
> >> request.  I'll review it for you ;)
> >>
> >>  - David
> >>
> >>
> >> On Tue, Jul 28, 2015 at 8:27 AM, David Thompson <
> david.thompson at kitware.com> wrote:
> >> Another project I work on has been using Bandit[1] to good effect.
> >>
> >>         David
> >>
> >> [1]:http://banditcpp.org/reference.html
> >>
> >> On Jul 28, 2015, at 10:01 AM, David Cole via vtk-developers <
> vtk-developers at vtk.org> wrote:
> >>
> >> > A single header file unit testing solution sounds like a good thing.
> >> >
> >> > ... but ... seems like we could also take advantage of a
> >> > well-established generally used unit testing framework (like gtest or
> >> > CppUnit) just as easily, and not clutter the VTK source with
> >> > yet-another-custom-thing-to-learn-about-VTKs-infrastructure.
> >> >
> >> > Would it be just as easy to add unit testing using an external unit
> >> > testing package? Or are there things about that scenario that are
> >> > overly complicated that having our own makes simpler...?
> >> >
> >> >
> >> > David C.
> >> >
> >> >
> >> >
> >> > On Mon, Jul 27, 2015 at 5:25 PM, David Gobbi <david.gobbi at gmail.com>
> wrote:
> >> >> Hi All,
> >> >>
> >> >> I've always enjoyed using the python unit testing framework in
> "vtk.test".
> >> >> It's been brought up on the list before, but it would be nice if VTK
> >> >> included a unit test framework for C++, as well.
> >> >>
> >> >> For my work projects, I've been using a unit test header that I
> wrote a year
> >> >> and a half ago:
> >> >> https://github.com/dgobbi/UnitTest/blob/master/UnitTest.h
> >> >>
> >> >> It's a single header file that provides a collection of macros for
> building
> >> >> unit tests, it's meant to be innocuous and easy to use.  It could go
> into
> >> >> VTK as Testing/Core/vtkUnitTest.h and people could use it at their
> leisure.
> >> >>
> >> >> Any thoughts?
> >> >>
> >> >> - David
> >>
> >>
> >
> > _______________________________________________
> > Powered by www.kitware.com
> >
> > Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
> >
> > Search the list archives at:
> http://markmail.org/search/?q=vtk-developers
> >
> > Follow this link to subscribe/unsubscribe:
> > http://public.kitware.com/mailman/listinfo/vtk-developers
> >
>
>
>
> --
> Unpaid intern in BillsBasement at noware dot com
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Search the list archives at: http://markmail.org/search/?q=vtk-developers
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/vtk-developers
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20150728/7e1a14a4/attachment.html>