[vtkusers] Compile Bug in VTK 5.6

Michael Jackson mike.jackson at bluequartz.net
Wed Jul 28 14:24:18 EDT 2010 

Actually Sean and I have been submitting some actual clang bugs to the  
clang developers over the last few days which they have fixed in  
record time. These clang bugs were holding up Sean's compiles. Now  
that those clang bugs are fixed the compiles should get much farther.  
I have been able to compile VTK with clang SVN HEAD as of today using  
shared libs and testing but NOT any type of wrapping. I should be  
submitting an experimental dashboard under the name "BlueQuartz" in a  
bit. I'll let Sean take over the rest of the testing duties after he  
gets his clang updated.
___________________________________________________________
Mike Jackson                      www.bluequartz.net

On Jul 28, 2010, at 2:18 PM, David Gobbi wrote:

> Sean McBride has an experimental clang dashboard that runs every
> night, but more would be better.  His dashboard is hung up on those
> USE_N_WAY_ARRAY clang errors so it doesn't provide info about anything
> else.  Other clang dashboards with different configurations would help
> to provide a lot more info.
>
>  David
>
>
> On Wed, Jul 28, 2010 at 12:13 PM, Michael Jackson
> <mike.jackson at bluequartz.net> wrote:
>> Dunno. I found them when I tried compiling VTK with the latest clang
>> compiler which is evidently pretty darn strict about checking  
>> everything.
>> ___________________________________________________________
>> Mike Jackson                      www.bluequartz.net
>> Principal Software Engineer       mike.jackson at bluequartz.net
>> BlueQuartz Software               Dayton, Ohio
>>
>>
>>
>> On Jul 28, 2010, at 2:05 PM, Marcus D. Hanwell wrote:
>>
>>> Hi,
>>>
>>> I will fix those files up, it looks like it was a simple error in  
>>> the
>>> declaration order that didn't produce a compilation error anywhere  
>>> that was
>>> tested. Sorry about that - is there a nightly dashboard this will/ 
>>> should be
>>> showing up on? Thanks for reporting these issues.
>>>
>>> Marcus
>>> --
>>> Marcus D. Hanwell, Ph.D.
>>> R&D Engineer, Kitware Inc.
>>> (518) 881-4937
>>>
>>> On Wed, Jul 28, 2010 at 1:59 PM, David Gobbi <david.gobbi at gmail.com>
>>> wrote:
>>> Hi Michael,
>>>
>>> I've pushed the vtkTIFFReader fix.  I'm forwarding bug
>>> http://public.kitware.com/Bug/view.php?id=11052 to Marcus because he
>>> probably has some idea about why that anonymous namespace is used in
>>> vtkPlotBar.cxx.
>>>
>>>  David
>>>
>>>
>>> On Wed, Jul 28, 2010 at 11:33 AM, Michael Jackson
>>> <mike.jackson at bluequartz.net> wrote:
>>>> I had just put a #if 0 ... #endif around the code because I was  
>>>> not sure
>>>> what the "correct" fix was. Simply deleting the functions will  
>>>> achieve
>>>> the
>>>> same thing.
>>>>
>>>> If you can take a look at the other bug it also proposes some  
>>>> fixes for
>>>> clang compile errors.
>>>>
>>>> Thanks
>>>> ___________________________________________________________
>>>> Mike Jackson                      www.bluequartz.net
>>>> Principal Software Engineer       mike.jackson at bluequartz.net
>>>> BlueQuartz Software               Dayton, Ohio
>>>>
>>>>
>>>>
>>>> On Jul 28, 2010, at 1:23 PM, David Gobbi wrote:
>>>>
>>>>> Did you try the fix that I suggested, i.e. just removing those two
>>>>> functions?  If that works, then I can commit the fix immediately.
>>>>>
>>>>> I should probably get clang, so that I can test code against it
>>>>> myself.  I'm not fond of xcode, but clang is probably going to  
>>>>> become
>>>>> a more and more common compiler as time goes on.
>>>>>
>>>>>  David
>>>>>
>>>>>
>>>>> On Wed, Jul 28, 2010 at 11:12 AM, Michael Jackson
>>>>> <mike.jackson at bluequartz.net> wrote:
>>>>>>
>>>>>> I filed bug http://public.kitware.com/Bug/view.php?id=11051 for  
>>>>>> this
>>>>>> issue
>>>>>> and http://public.kitware.com/Bug/view.php?id=11052 for a related
>>>>>> issue.
>>>>>> I
>>>>>> don't have git commit access so I am hoping someone else can do  
>>>>>> that
>>>>>> part.
>>>>>>
>>>>>> I'll hopefully be submitting an experimental dashboard within  
>>>>>> the hour
>>>>>> with
>>>>>> the complete compile/test results for VTK.
>>>>>>
>>>>>> I think I was on the actual VTK 5.6 release from GIT.
>>>>>> ___________________________________________________________
>>>>>> Mike Jackson                      www.bluequartz.net
>>>>>>
>>>>>>
>>>>>> On Jul 28, 2010, at 1:04 PM, David Gobbi wrote:
>>>>>>
>>>>>>> Do you mean VTK git master?  There's a difference between VTK  
>>>>>>> 5.6 and
>>>>>>> the latest VTK from git.
>>>>>>>
>>>>>>> I think that vtkTIFFReader.cxx has some old templated functions
>>>>>>> inside
>>>>>>> of it that aren't used anymore.  The right way to patch the file
>>>>>>> would
>>>>>>> be to simply remove those functions (ReadTiledImage and
>>>>>>> ReadScanlineImage) completely.
>>>>>>>
>>>>>>>  David
>>>>>>>
>>>>>>>
>>>>>>> On Wed, Jul 28, 2010 at 10:23 AM, Michael Jackson
>>>>>>> <mike.jackson at bluequartz.net> wrote:
>>>>>>>>
>>>>>>>> I am getting the following compiler error for VTK 5.6
>>>>>>>>
>>>>>>>> [ 57%] /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.cxx: 
>>>>>>>> 811:19:
>>>>>>>> error:
>>>>>>>> 'EvaluateImageAt' is a protected
>>>>>>>>    member of 'vtkTIFFReader'
>>>>>>>>          self->EvaluateImageAt(image+imagepos,
>>>>>>>> static_cast<T*>(buffer)+tilepos);
>>>>>>>>                ^
>>>>>>>> In file included from
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.cxx:15:
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.h:116:7: note:
>>>>>>>> declared
>>>>>>>> protected here
>>>>>>>>  int EvaluateImageAt( void*, void* );
>>>>>>>>    ^
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.cxx:829:47:  
>>>>>>>> error:
>>>>>>>> 'GetInternalImage' is a protected
>>>>>>>>    member of 'vtkTIFFReader'
>>>>>>>>  unsigned int isize =
>>>>>>>> TIFFScanlineSize(self->GetInternalImage()->Image);
>>>>>>>>                                            ^
>>>>>>>> In file included from
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.cxx:15:
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.h:114:26: note:
>>>>>>>> declared
>>>>>>>> protected here
>>>>>>>>  vtkTIFFReaderInternal *GetInternalImage() { return
>>>>>>>> this->InternalImage;
>>>>>>>> }
>>>>>>>>                       ^
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.cxx:852:21:  
>>>>>>>> error:
>>>>>>>> 'EvaluateImageAt' is a protected
>>>>>>>>    member of 'vtkTIFFReader'
>>>>>>>>      inc = self->EvaluateImageAt( image,
>>>>>>>>                  ^
>>>>>>>> In file included from
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.cxx:15:
>>>>>>>> /Users/Shared/OpenSource/VTK/IO/vtkTIFFReader.h:116:7: note:
>>>>>>>> declared
>>>>>>>> protected here
>>>>>>>>  int EvaluateImageAt( void*, void* );
>>>>>>>>    ^
>>>>>>>>
>>>>>>>> Looking at where those errors occur reveals that they are in
>>>>>>>> template
>>>>>>>> functions inside of ReadTiledImage and ReadScanlineImage in the
>>>>>>>> vtkTIFFReader.cxx file. These functions appear never to be used
>>>>>>>> anywhere
>>>>>>>> in
>>>>>>>> the code but if they _do_ get used then a compile error will be
>>>>>>>> thrown
>>>>>>>> at
>>>>>>>> that point. The functions probably need to be moved to the
>>>>>>>> vtkTIFFReader.h
>>>>>>>> file instead of in the vtkTIFFReader.cxx file or the  
>>>>>>>> functions need
>>>>>>>> to
>>>>>>>> be
>>>>>>>> removed if they are not being used at all.
>>>>>>>>
>>>>>>>>  If the vtk-devs are ok with that I can submit a patch for  
>>>>>>>> someone
>>>>>>>> to
>>>>>>>> commit.
>>
>>

More information about the vtkusers mailing list