[Insight-developers] itkFactoryTestLib.cxx & Sun CC

Mon Apr 6 02:17:29 EDT 2009

Hi Luis,

Thank you so much for handling this. I surely didn't expect all the
required changes with respect to this patch.

Cheers,
Tom

On Fri, Apr 3, 2009 at 21:52, Luis Ibanez <luis.ibanez at kitware.com> wrote:
> Hi Tom,
>
> I'm still making changes in the Sun build.
> Probably won't be able to finish before the Dashboar rollup tonight.
> I'll continue with this tomorrow.
>
> It is almost tempting to simply create a class deriving from the std::allocator
> and to implement the destroy() method on it...
>
> BTW, In the process I came across some VXL symbols:
>
> grep -r SUNPRO * | grep ALL
>
>
> ~/src/Insight/Utilities/vxl
> vcl/vcl_compiler.h:# define VCL_SUNPRO_ALLOCATOR_HACK(T) T
> VCL_SUNPRO_CLASS_SCOPE_HACK(std::allocator<T >)
> vcl/vcl_compiler.h:# define VCL_SUNPRO_ALLOCATOR_HACK(T) T // FIXME
> vcl/sunpro/vcl_map.h:// this avoids the VCL_SUNPRO_ALLOCATOR_HACK
> vcl/sunpro/vcl_vector.h:// this avoids the VCL_SUNPRO_ALLOCATOR_HACK
> vcl/emulation/vcl_hashtable.h:  typedef
> vcl_vector<VCL_SUNPRO_ALLOCATOR_HACK(node*) > buckets_type;
>
>
> that lead to suspect that the VXL project may have already
> dealt with this problem... (not sure if that's the case).
>
>
>       Luis
>
> -------------------------------------------------------------
> On Fri, Apr 3, 2009 at 2:52 PM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>> Hi Tom
>>
>> Here is what I have found so far:
>>
>> A) Using the -D_RWSTD_ALLOCATOR
>>     symbol resulted in other compilation errors.
>>
>>
>> B) Using the compiler flag:
>>
>>                 -library=stlport4
>>
>>  seems to work fine, but it requires that we add
>>  a couple of standard includes in several files.
>>  ( #include <string.h>, <stdlib.h>...)
>>
>>  This doesn't seem to be much of a problem...
>>
>>  Attached is the patch so far...
>>
>>  It is still compiling in this Sun station...
>>  once it finishes, I'll try the patch on Linux,
>>  and if clean, will commit it.
>>
>>
>>       Luis
>>
>>
>>
>> ---------------------------
>> On Fri, Apr 3, 2009 at 10:26 AM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>>> Hi Tom,
>>>
>>> Thanks for tracking the error in the dashboard.
>>>
>>> From what I understand from the forum post that
>>> you found, it seems that adding the compiler flag:
>>>
>>>            -D_RWSTD_ALLOCATOR
>>>
>>> would be the best option.
>>>
>>> Normally we avoid adding this type of defines in the
>>> compiler command line, and we rather set the #defines
>>> in the file.
>>>
>>>                         itkConfigure.h.in
>>>
>>> With a line similar to:
>>>
>>>       #cmakedefine   _RWSTD_ALLOCATOR
>>>
>>> Let me try doing this change in this Sun machine (it is at Kitware),
>>> and will report later on how it goes.
>>>
>>>
>>>     Thanks,
>>>
>>>
>>>          Luis
>>>
>>>
>>> ----------------------------------
>>> On Fri, Apr 3, 2009 at 9:06 AM, Tom Vercauteren <tom.vercauteren at m4x.org> wrote:
>>>> Hi,
>>>>
>>>> I am going to need help on that one:
>>>> http://www.cdash.org/CDash/viewBuildError.php?buildid=305373
>>>>
>>>> Sun cc compiler seems to not like the std::allocator :(
>>>>
>>>> I should be able to fix the "Cannot assign void* to XXX*" easily by
>>>> simply adding an explicit static_cast.
>>>>
>>>> However I don't know how to handle the "destroy is not a member of
>>>> std::allocator" issue. The only interesting information I found is
>>>> here:
>>>>  http://forums.sun.com/thread.jspa?threadID=5317335
>>>> where two possible compiler options are proposed to fix the problem:
>>>>  -library=stlport4
>>>> or
>>>>  -D_RWSTD_ALLOCATOR
>>>>
>>>> Having absolutely no experience of sun, I have no clue whether any of
>>>> those options seems reasonable for ITK.
>>>>
>>>> Another option of course would be to drop the use of std::allocator
>>>> either entirely or only for sun within that unit test. However I do
>>>> think that this std::allocator provides a nice example on how to
>>>> extend ITK.
>>>>
>>>> Thoughts?
>>>>
>>>> Tom
>>>>
>>>> On Thu, Apr 2, 2009 at 19:53, Tom Vercauteren <tom.vercauteren at m4x.org> wrote:
>>>>> Thanks Bill,
>>>>>
>>>>> Even though, on that page
>>>>>  http://msdn.microsoft.com/en-us/library/723te7k3.aspx
>>>>> the prototype of allocate is shown as requiring 2 arguments, the
>>>>> example provided on the same page only uses one.
>>>>>
>>>>> Also, MSVC 2003 has a default argument for the second argument:
>>>>>  http://msdn.microsoft.com/en-us/library/723te7k3(VS.71).aspx
>>>>>
>>>>> Anyhow, let's see how it goes.
>>>>>
>>>>> Tom
>>>>>
>>>>> On Thu, Apr 2, 2009 at 19:42, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>>>>>> It's not just VS6. See for example:
>>>>>> http://msdn.microsoft.com/en-us/library/723te7k3.aspx
>>>>>> which is for VS 2008.
>>>>>>
>>>>>> Your fix should be portable.
>>>>>>
>>>>>> Bil
>>>>>>
>>>>>> On Thu, Apr 2, 2009 at 10:35 AM, Tom Vercauteren
>>>>>> <tom.vercauteren at m4x.org> wrote:
>>>>>>> Hey Luis,
>>>>>>>
>>>>>>> It seems that this is a known msvc 6.0 bug. See chapter 11 of
>>>>>>> http://www.acceleratedcpp.com/details/msbugs.html
>>>>>>>
>>>>>>> I'll try with adding a dummy second argument (null pointer). I hope it
>>>>>>> won't make other platforms fail...
>>>>>>> http://www.itk.org/cgi-bin/viewcvs.cgi/Testing/Code/Common/itkFactoryTestLib.cxx?root=Insight&r1=1.5&r2=1.6&sortby=date
>>>>>>>
>>>>>>> Tom
>>>>>>>
>>>>>>> On Thu, Apr 2, 2009 at 18:42, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>>>>>>>>
>>>>>>>> Hi Tom,
>>>>>>>>
>>>>>>>> It seems that Visual Studio 6.0 has difficulty
>>>>>>>> with the changes in itkFactoryTestLib.cxx:
>>>>>>>>
>>>>>>>> http://www.cdash.org/CDash/viewBuildError.php?buildid=304766
>>>>>>>>
>>>>>>>>
>>>>>>>> Compiling...
>>>>>>>> itkFactoryTestLib.cxx
>>>>>>>>
>>>>>>>> c:\Dashboards\My
>>>>>>>> Tests\InsightContinuous\Testing\Code\Common\itkFactoryTestLib.cxx(68) :
>>>>>>>> error C2660: 'allocate' : function does not take 1 parameters
>>>>>>>>
>>>>>>>>        c:\Dashboards\My
>>>>>>>> Tests\InsightContinuous\Testing\Code\Common\itkFactoryTestLib.cxx(59) :
>>>>>>>> while compiling class-template member function 'short *__thiscall
>>>>>>>> TestImportImageContainer<unsigned long,short>::AllocateElements(unsigned
>>>>>>>> long) const'
>>>>>>>>
>>>>>>>>
>>>>>>>>  Any suggestions ?
>>>>>>>>
>>>>>>>>
>>>>>>>>     Thanks
>>>>>>>>
>>>>>>>>
>>>>>>>>         Luis
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> 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
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>