Re: [Swig-user] world's longest wrapper compile-time warnings with MS VC7

[William S F. <ws...@fu...>]

Nitro wrote:
> Am 14.09.2006, 23:02 Uhr, schrieb James Carroll <mr...@gm...>:
> 
>> I've got a swig c++ python project (swig 1.3.27) that works pretty
>> well from a python setup.py build envirnoment, and I'm putting
>> togteher a VC7 build project, roughly following:
>>   http://www.geocities.com/foetsch/python/extending_python.htm
>>
>> I'm creating a dll, and then also creating a second dll for my python
>> wrapper which just has the swig-generated .cpp.
>>
>> I'm getting warnings that look like:
>>
>> C:\Program Files\Microsoft Visual Studio .NET
>> 2003\Vc7\include\iterator(324) : warning C4244: 'argument' :
>> conversion from 'std::vector<_Ty>::iterator::difference_type' to
>> 'swig::PySequence_Iter<T,Reference>::difference_type', possible loss
>> of data
>>         with
>>         [
>>             _Ty=mbf::Datanode
>>         ]
>>         and
>>         [
>>             T=swig::traits_asptr_stdseq<std::vector<mbf::Datanode>>::value_type,
>>             Reference=const
>> swig::PySequence_Cont<swig::traits_asptr_stdseq<std::vector<mbf::Datanode>>::value_type>::const_reference
>>         ]
>> ...  (the whole warning is in the attached txt file.)
>>
>>
>> What am I doing to create the conflict in difference_type classes?
> 
> It's not you, it's the swig code :) I get those myself, too. I think some  
> of them were fixed in swig 1.3.29. But there are still some that persist.  
> I know they can be a pain because a single warning takes 200 lines in the  
> log. Maybe you should try upgrading to 1.3.29 and see if your warnings  
> persist. If yes, see which code causes this problem and try to find it in  
> the swig sources (simple search in files should suffice). Then report back  
> here, so that the swig devs can incorporate this back into swig.
> 

Also you can use the /WL command line option so that Visual C++ only
generates one line per warning (a very long line, but still one line).

William