[Insight-developers] MeshIO: Low coverage uncovers errors

[wanlin]

Hi, Bill
    First i am sorry that i didn't read the new code contribution policy
before i push the patch to ITK gerrit. To be honest, only the VTK legacy
file format was used regularly so i provide more tests (Ascii, Binary, with
point normals, with point scalar). I should have comprehensive tests and
conduct converge tests for all the supported formats as well. I will improve
the patch later. Sorry for this.


Kind regards

wanlin

On Sat, Sep 3, 2011 at 6:50 AM, Bill Lorensen <bill.lorensen at gmail.com>wrote:

> Folks,
>
> I am concerned about the lack of test coverage in newly merged code. I
> decided to look at the newly merged MeshIO classes.
>
> Topic: http://review.source.kitware.com/#change,2422
>
> The topic had 7 reviewers (I was one of them).  It seems that it was
> approved because it compiled and all of the author's tests passed. However,
> the code coverage is very low. I never looked at the code before the merge
> (my bad).
>
> I looked at the OBJMeshIO class, since I am familiar with the Wavefront obj
> format. The class had very low coverage. I modified the input file so that
> it contained point normals. The code failed to parse the file properly. The
> current topic parses the file properly, but does not populate the mesh point
> data correctly.
>
> New topic: http://review.source.kitware.com/#change,2641
>
> I am concerned that the MeshIO classes have not been validated.
>
> This illustrates a shortcoming in our current process. Code is getting
> merged that has low test coverage. Also, some results are difficult to
> validate.
>
> I will continue to look at these classes and fix the formats that I am
> familiar with. I hope others can help.
>
> Bill
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20110903/d00d3f51/attachment.htm>