[Insight-developers] Adding Concept Chechking "float"/"double" to pixel type of GradientAnisotropicDiffusion

[Bill Lorensen]

Steve,

According to our backward compatibility policy, the introduction of compiler
errors should be avoided at all costs. A compiler error is typically crytic
especially in templated code. A runtime error message can be very
descriptive to the owner of the code. As you stated, the development team
failed to place this restriction on the code originally. Often, the person
recompiling the application code may not be the same person who wrote the
original application. Also, some time may pass between version updates of
third p[arty (e.g. ITK, VTK) code.

Granted, there will be times when a compile error cannot be avoided, but
this should be a last resort.

I was just going to quote the ITK Bacrwd Compatibility document, but I'm
emrassed to say, I can't find it on the internet. I'll track it down...

Bill

On 8/15/07, Steve M. Robbins <steve at sumost.ca> wrote:
>
> > On 8/15/07 8:12 AM, "Bill Lorensen" <bill.lorensen at gmail.com> wrote:
> >
> > > Luis,
> > >
> > > This change may cause backward compatibility problems. Old code may no
> longer
> > > compile and the compiler error for the conept checking is cryptic on
> some
> > > compilers. We should warn the user and give explicit instructions on
> how to
> > > repair the code in the warning message.
> > >
> > > I understand that the current code is not producing correct results.
> But that
> > > is the fault of the itk development team and not the user of the code.
> We
> > > should notify users in an instructive way on how to fix their code so
> that it
> > > produces correct results. A cryptic compiler error is not informative.
>
> For my curiousity only: how is it a failure of the ITK development team?
> Naively, it seems the failure is not to have had the concept checking
> from Day 1.  Or did you have something else in mind?
>
> For what it's worth (not being an ITK developer and never having seen
> the fallout of a concept check failure), I tend to agree with Luis.
> IMHO, it is better to have the build fail than the execution because
> the user may not be in the position to fix the error, assuming they
> see the message at all.
>
> Regards,
> -Steve
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
>
> iD8DBQFGwyZN0i2bPSHbMcURAtAaAJ4qVgiuaetMmebNEezc2JIxg8qzeACeJt6Q
> xH09gWNN53ryTkubhSukdxI=
> =lyp8
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at itk.org
> http://www.itk.org/mailman/listinfo/insight-developers
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.itk.org/mailman/private/insight-developers/attachments/20070815/c548c8b5/attachment.htm