[Insight-developers] Insight-developers Digest, Vol 61, Issue 31

[Bill Lorensen]

I think it will take a couple of months to root out all of the bad
behaviour that we released in the past. We will warn users that the
portable rounding and centered pixel code may change. I think this
will avoid adding addition backward compatibility issues.

I agree that these are important changes. Unfortunately there is not a
way do do this with the Review mechanism. We'll have to ask our
customers to accept the new release with the old, admittedly,
inconsistent, behaviour, while we do our best to present a correct
solution.

Bill

On Mon, May 25, 2009 at 3:09 PM, Simon Warfield
<simon.warfield at childrens.harvard.edu> wrote:
> Bill Lorensen wrote:
>>
>> Simon,
>>
>> We still have not validated the portable rounding. Having it available
>> during the upcoming release will give us a chance to make sure
>> everything is correct before we make it the default. I suspect that
>> once the portable rounding is validated we can do away with the old,
>> incorrect way.
>>
>
> I think the key will be to identify the other parts of the code base that
> don't work correctly with portable rounding in conjunction with a consistent
> physical coordinate system,
> and to move them to work correctly as quickly as possible.
>
> That will involve understanding how the regression test results change, and
> selecting the new test results which represent correct operation.
> That is currently how validation of code changes are made. There isn't a
> separate validation process.
>
> Perhaps it would be better to hold off on the release, rather than incur an
> additional backward compatibility burden.
>
> --
> Simon
>
>> Bill
>>
>> On Mon, May 25, 2009 at 2:49 PM, Simon Warfield
>> <simon.warfield at childrens.harvard.edu> wrote:
>>
>>>
>>> Dear Luis,
>>>
>>> Either option A or B make sense to me.
>>>
>>> At the moment, the patch to fix the bug of non- portable rounding,  which
>>> is
>>> critical for indexes to be calculated the same way across platforms, is
>>> not
>>> fully satisfactory.
>>> The current patch ensures inconsistent cross-platform behavior if
>>> ITK_USE_PORTABLE_ROUND is not selected.
>>> This would be appropriate if we were selecting to turn on or off some
>>> feature, but here it is being used to turn on or off correct behavior.
>>>
>>>>
>>>> Let me propose two options:
>>>>
>>>>
>>>>
>>>>   A) We commit your patch, and post-pone the release
>>>>      until next Saturday.
>>>>
>>>>      This will give us time to verify the all builds
>>>>      are correct
>>>>
>>>>
>>>>
>>>>   B) We hold on the patch. Cut the release as scheduled.
>>>>      and commit the patch immediately after the repository
>>>>      reopens.
>>>>
>>>>      Once the dashboard stabilizes, we cut a patch release
>>>>      ITK 3.14.1.
>>>>
>>>>
>>>>
>>>>
>>>> If we are going to proceed with (A),
>>>> we should at least see an Experimental build first.
>>>>
>>>>
>>>>
>>>>   What other developers think ?
>>>>
>>>>
>>>>      Luis
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>> --
>>> Simon
>>>
>>>
>>> _______________________________________________
>>> Powered by www.kitware.com
>>>
>>> Visit other Kitware open-source projects at
>>> http://www.kitware.com/opensource/opensource.html
>>>
>>> Please keep messages on-topic and check the ITK FAQ at:
>>> http://www.itk.org/Wiki/ITK_FAQ
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://www.itk.org/mailman/listinfo/insight-developers
>>>
>>>
>>
>>
>
>
> --
> Simon
>
>
>