[Insight-developers] Problems compiling today

[Bradley Lowekamp]

Bill,

It looks like my solution was less correct :(

As we are getting a bunch of compilation error on windows now:

http://open.cdash.org/viewBuildError.php?buildid=2848943

What concerned me about your approach was that is blanketly just changes the export specification for all unspecified methods and classes. So it likely will have un intended consequences sooner or later.

Perhaps we should just add some "ifdef GCC"s around the statements I added? Better yet define "ITK_EXPLICIT_SPECIALIZATION_REQUIRES_EXPORT" in the config...

Brad


On Mar 16, 2013, at 10:39 AM, Bradley Lowekamp <blowekamp at mail.nih.gov> wrote:

> Bill,
> 
> Looks like you caught a major issue right before the dashboard did :) You have impeccable timing!
> 
> I looked at this. And it  is just a case of a missing ITKCommon_Export. It's just a little tricky because you have the non-specilized declaration and definition. Then a specialized declaration. What's missing is a statement which says the method has been explicitly instantiated (this just happens to be a totally specialized version), but the statement is that same as if it'd been explicitly instantiated. This has been compiler dependent, and we use the ITKCommon_EXPORT macro for that. It's sounds like the extern keyword should to this in C++0x, but I hadn't looked at the closely.
> 
> I hope this help explains what's going on....
> 
> Brad
> 
> 
> On Mar 15, 2013, at 10:31 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
> 
>> There are issues with template specializations and -fvisibility=hidden. From what I read no current compilers officially support this, but most do. I guess my gcc on the MAXC does not.
>> 
>> I'll put a patch to gerrit that adds lines like these in Common/src/CMakeLists.txt
>>     if(APPLE)
>>       set_source_files_properties( itkArrayOutputSpecialization.cxx PROPERTIES
>>         COMPILE_FLAGS -fvisibility=default )
>>     endif()
>> 
>> When I do this, my gcc build is OK.
>> 
>> Bill
>> 
>> On Fri, Mar 15, 2013 at 7:00 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>> It has to do with symbol visibility. If I do nm on the ITKCommon lib, the undefined symbols have "t" and not "T". "t" means the symbols are local. If I remove the -fvisiibility=hidden flag from ITKCommon's CMakeLists.txt file, the tests build fine.
>> 
>> 
>> 
>> On Fri, Mar 15, 2013 at 6:49 PM, Bradley Lowekamp <blowekamp at mail.nih.gov> wrote:
>> Bill,
>> 
>> Are you able to compile the release? Or is it a  change since then?
>> 
>> Brad
>> 
>> On Mar 15, 2013, at 7:36 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>> 
>>> I see why  those statements are there.
>>> 
>>> Still need help.
>>> 
>>> On Fri, Mar 15, 2013 at 4:25 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>>> More info. If I remove
>>> template<> std::ostream & operator<< <double> (std::ostream & os, const Array< double > & arr);
>>> template<> std::ostream & operator<< <float> (std::ostream & os, const Array< float > & arr);
>>> 
>>> Both my c++ and clang build fine.
>>> 
>>> What is the purpose of these lines?
>>> 
>>> Bill
>>> 
>>> On Fri, Mar 15, 2013 at 4:09 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>>> Guys,
>>> 
>>> This is  on my Mac using:
>>> c++ --version
>>> c++ --version
>>> i686-apple-darwin11-llvm-g++-4.2 (GCC) 4.2.1 (Based on Apple Inc. build 5658) (LLVM build 2336.1.00)
>>> 
>>> My clang version builds fine.
>>> 
>>> Any ideas?
>>> 
>>> Bill
>>> 
>>> 
>>> On Fri, Mar 15, 2013 at 11:52 AM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
>>> Folks,
>>> 
>>> It's been a few days since itk and tried a build.
>>> 
>>> Today I get Linking CXX executable ../../../../bin/ITKCommon1TestDriver
>>> Undefined symbols for architecture x86_64:
>>>   "std::basic_ostream<char, std::char_traits<char> >& itk::operator<< <float>(std::basic_ostream<char, std::char_traits<char> >&, itk::Array<float> const&)", referenced from:
>>>       void CheckVariableLengthArrayTraits<itk::Array<float> >(itk::Array<float>)in itkNumericTraitsTest.cxx.o
>>>       itkArrayTest(int, char**)in itkArrayTest.cxx.o
>>>   "std::basic_ostream<char, std::char_traits<char> >& itk::operator<< <double>(std::basic_ostream<char, std::char_traits<char> >&, itk::Array<double> const&)", referenced from:
>>>       void CheckVariableLengthArrayTraits<itk::Array<double> >(itk::Array<double>)in itkNumericTraitsTest.cxx.o
>>> ld: symbol(s) not found for architecture x86_64
>>> collect2: ld returned 1 exit status
>>> make[2]: *** [bin/ITKCommon1TestDriver] Error 1
>>> make[1]: *** [Modules/Core/Common/test/CMakeFiles/ITKCommon1TestDriver.dir/all] Error 2
>>> make: *** [all] Error 2
>>> 
>>> 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
>>> _______________________________________________
>>> 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.php
>>> 
>>> 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
>> 
>> 
>> 
>> 
>> -- 
>> Unpaid intern in BillsBasement at noware dot com
>> 
>> 
>> 
>> -- 
>> Unpaid intern in BillsBasement at noware dot com
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-developers/attachments/20130318/f43b5d28/attachment.htm>