[Insight-developers] Video and OpenCV issues

Luis Ibanez luis.ibanez at kitware.com
Fri Nov 4 10:36:50 EDT 2011


Bill,

This was mostly due to our learning process on OpenCV...

We probably should review the code and replace the
use of the C interface with the more modern C++ interface.


    Luis


---------------------
On Thu, Nov 3, 2011 at 3:00 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
> Eventually, you will see that cvCaptureFromFile is failing in  CanReadFile
>
> BTW, why is the openCV C interface being used rather than the C++interface?
>
> On Thu, Nov 3, 2011 at 2:48 PM, Bill Lorensen <bill.lorensen at gmail.com>
> wrote:
>>
>> I pulled down the patch and still get these errors:
>>
>>     Start 1: OpenCVVideoCaptureTest
>> 1/4 Test #1: OpenCVVideoCaptureTest ...........***Exception: SegFault
>> 1.44 sec
>>     Start 2: OpenCVImageBridgeTest
>> 2/4 Test #2: OpenCVImageBridgeTest ............   Passed    0.95 sec
>>     Start 3: OpenCVVideoIOTest
>> 3/4 Test #3: OpenCVVideoIOTest ................***Failed    0.25 sec
>>     Start 4: OpenCVVideoIOFactoryTest
>> 4/4 Test #4: OpenCVVideoIOFactoryTest .........***Failed    0.26 sec
>>
>> The code is not checking for failure here. It shopuld throw an exception.
>> Of course this will not fix the problem.
>>
>> 247    ::InitializeVideoIO()
>> 248    {
>> 249      m_VideoIO = itk::VideoIOFactory::CreateVideoIO(
>> 250                                    itk::VideoIOFactory::ReadFileMode,
>> 251                                    m_FileName.c_str());
>> 252      m_VideoIO->SetFileName(m_FileName.c_str());
>> 253      m_VideoIO->ReadImageInformation();
>> 254
>>
>>
>> On Thu, Nov 3, 2011 at 2:15 PM, Luis Ibanez <luis.ibanez at kitware.com>
>> wrote:
>>>
>>> Bill,
>>>
>>> We made some progress with Xiaoxiao,
>>> and push a patch 4:
>>>
>>> Where the ITK_VIDEO_USE_*
>>> variables were part of the problem,
>>>
>>> http://review.source.kitware.com/#patch,sidebyside,2997,4,Modules/Video/IO/src/itkVideoIOFactory.cxx
>>>
>>> but,...
>>> still need to do further debugging.
>>>
>>> We think that we are getting closer
>>> to the final source of the problem.
>>>
>>>
>>>     Luis
>>>
>>>
>>> -----------------------------------------------
>>> On Thu, Nov 3, 2011 at 1:27 PM, Bill Lorensen <bill.lorensen at gmail.com>
>>> wrote:
>>> > I built my OpenCV with FFMPEG support. I assume you'll find the same
>>> > problem
>>> > when I ran gdb.
>>> >
>>> > Quoted from my gerrit review:
>>> >
>>> > OpenCVVideoIO::CanReadFile, cvCaptureFromFile fails.VideoFileReader
>>> > should
>>> > check for failure here:
>>> > m_VideoIO = itk::VideoIOFactory::CreateVideoIO(
>>> > itk::VideoIOFactory::ReadFileMode,
>>> > m_FileName.c_str());
>>> > m_VideoIO->SetFileName(m_FileName.c_str());
>>> > m_VideoIO->ReadImageInformation();
>>> >
>>> > On Thu, Nov 3, 2011 at 1:12 PM, Luis Ibanez <luis.ibanez at kitware.com>
>>> > wrote:
>>> >>
>>> >> Bill,
>>> >>
>>> >> I just managed to replicate the segmentation fault in:
>>> >>
>>> >>              OpenCVVideoCaptureTest
>>> >>
>>> >>
>>> >> ... now firing up GDB...    :-)
>>> >>
>>> >>
>>> >> I would suspect that this is related to not having build
>>> >> some FFMPEG support in OpenCV, but, I will probably
>>> >> be wrong and will end up having to buy you a beer...
>>> >>
>>> >>
>>> >>     Luis
>>> >>
>>> >>
>>> >> ------------------------------------------------------------
>>> >> On Fri, Oct 28, 2011 at 11:42 AM, Bill Lorensen
>>> >> <bill.lorensen at gmail.com>
>>> >> wrote:
>>> >> > Luis and Pat,
>>> >> >
>>> >> > I am building openvc 2.3.1 with WITH_FFMPEG:BOOL=ON.
>>> >> >
>>> >> > This is against this gerrit topic:
>>> >> > http://review.source.kitware.com/#change,2997
>>> >> >
>>> >> > itkOpenCVVideoCaptureTest segfaults.
>>> >> >
>>> >> > Bill
>>> >> >
>>> >
>>> >
>>> >
>>> > --
>>> > Unpaid intern in BillsBasement at noware dot com
>>> >
>>> >
>>
>>
>>
>> --
>> Unpaid intern in BillsBasement at noware dot com
>>
>
>
>
> --
> Unpaid intern in BillsBasement at noware dot com
>
>


More information about the Insight-developers mailing list