[Insight-developers] Recent changes to arrow code are causing numerical exceptions

[Bill Lorensen]

Gabe,

No problem. We appreciate your efforts. All of us make these mistakes.
The important thing is to fix them as quickly as possible.

The VS6 problem is because that compiler is pretty broken. You will
either have to use an enum or a #define.

This page describes the issue:
http://support.microsoft.com/kb/241569

Bill

On Mon, Jun 14, 2010 at 9:20 AM, Gabe Hart <gabe.hart at kitware.com> wrote:
> Hi Bill,
>
> This has been my "learn the ropes" coverage/bug fix project, so I apologize
> for the messy dashboard the last few days.  The bug you found in the test is
> an artifact of trying to get things fixed before leaving on Friday
> (something Luis has just informed me is not probably a good thing to do).
>  The errors that were arising last week came about because I was trying to
> modify MetaArrowConverter so that it properly handled the Position and
> Direction fields from itkArrowSpatialObject.  My initial attempt at this was
> to put the Direction field from the spatial object into the Orientation
> field from the meta object.  After some investigation we realized that this
> was trying to shoehorn a vector into a matrix.  We then decided that the
> best course of action was adding a new field to the MetaArrow object to
> specifically hold the direction vector.  The bug you identified was a
> leftover from the first fix (Direction -> Orientation) which I should have
> caught when making the initial changes.  I'm fixing it now, so it should be
> all set soon.
>
> There does seem to be another issue which is coming from the VS6 compilers
> around initializing a constant in the MetaArrow.h header file.  I'll get on
> fixing this asap also.
>
> -Gabe
>
> On 06/14/2010 09:02 AM, Bill Lorensen wrote:
>>
>> OK, here is the problem I think.
>>
>> The zero divide is occurring in itkMetaArrowConverterTest at the
>> newMetaDirectionNorm.Normalize();
>> The newMetaOrientation is all zeroes...
>> Looking back, the direction is obtained from newMetaArrow which is
>> created by the converter.
>>
>> The converter internally creates a smartpointer to a spatialobject and
>> returns it. I believe this is not correct. Doesn't the smartpointer go
>> away when it loses scope?
>>
>> Looks like the code has been there for 5 1/2 years. Looking at an old
>> coverage log:
>>
>> http://www.cdash.org/CDash/viewCoverageFile.php?buildid=623555&fileid=10859113
>> seems like the MetaArrowConverter class was never tested.
>>
>> Bill
>>
>>
>> On Mon, Jun 14, 2010 at 7:40 AM, Luis Ibanez<luis.ibanez at kitware.com>
>>  wrote:
>>
>>>
>>> Hi Gabe,
>>>
>>> Can you please take a look at this problem ?
>>>
>>> A common cause is to have divisions by zero...
>>> (but there may be other reasons..).
>>>
>>>
>>>   Thanks
>>>
>>>
>>>       Luis
>>>
>>>
>>>
>>> ---------------------------------------------------------------------------
>>> On Sun, Jun 13, 2010 at 11:53 AM, Bill Lorensen<bill.lorensen at gmail.com>
>>> wrote:
>>>
>>>>
>>>> Folks,
>>>>
>>>> I did not try to track down the culprit, but recent changes to the
>>>> spatial object arrow code is causing numerical exceptions on the
>>>> borland compiler.
>>>>
>>>>
>>>> http://www.cdash.org/CDash/testSummary.php?project=2&name=itkMetaArrowConverterTest&date=2010-06-13
>>>>
>>>> This is probably not Borland specific. The Borland compiler is the
>>>> only one that enables floating point exceptions.
>>>>
>>>> Bill
>>>> _______________________________________________
>>>> Powered by www.kitware.com
>>>>
>>>> Visit other Kitware open-source projects at
>>>> http://www.kitware.com/opensource/opensource.html
>>>>
>>>> Kitware offers ITK Training Courses, for more information visit:
>>>> http://kitware.com/products/protraining.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
>>>>
>>>
>>>
>
>