[Pythonmac-SIG] wxPython build issue

[Kevin Ollivier]

Hi Bill,

On Feb 3, 2006, at 3:57 PM, Bill Northcott wrote:

>>> PS  I still have the issue from my other posting which Adriano  
>>> posted in September.
>>
>> Could you give me a thread to look up so that I can follow up?
>
> I spent a bit more time on this, but I have to quit now.
>
> The problem here is that if <math.h> is included with  
> _POSIX_C_SOURCE defined, the gamma(), rinttol() and roundtol()  
> functions among others are not declared.  wx/mac/private.h, which  
> is included from helpers.cpp, includes <Carbon/carbon.h> .  The  
> carbon header eventually includes fp.h.  The #ifdef on line 124 of  
> fp.h prevents the declarations of gamma() etc being used from  
> fp.h.  Hence the errors:
>
> gcc -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno- 
> fused-madd -fPIC -fno-common -dynamic -DNDEBUG -g -O3 -Wall - 
> Wstrict-prototypes -DSWIG_TYPE_TABLE=_wxPython_table - 
> DHAVE_CONFIG_H -DWXP_USE_THREAD=1 -UNDEBUG -D__WXMAC__ - 
> D_FILE_OFFSET_BITS=64 -D_LARGE_FILES -DNO_GCC_PRAGMA -Iinclude - 
> Isrc -I/usr/local/lib/wx/include/mac-ansi-release-2.6 -I/usr/local/ 
> include/wx-2.6 -I/Library/Frameworks/Python.framework/Versions/2.4/ 
> include/python2.4 -c src/helpers.cpp -o build/temp.darwin-8.4.0- 
> Power_Macintosh-2.4/src/helpers.o -O3
> cc1plus: warning: command line option "-Wstrict-prototypes" is  
> valid for C/ObjC but not for C++
> /System/Library/Frameworks/CoreServices.framework/Frameworks/ 
> CarbonCore.framework/Headers/fp.h: In function 'long double gammal 
> (long double)':
> /System/Library/Frameworks/CoreServices.framework/Frameworks/ 
> CarbonCore.framework/Headers/fp.h:2031: error: 'gamma' was not  
> declared in this scope
> /System/Library/Frameworks/CoreServices.framework/Frameworks/ 
> CarbonCore.framework/Headers/fp.h: In function 'long int rinttoll 
> (long double)':
> /System/Library/Frameworks/CoreServices.framework/Frameworks/ 
> CarbonCore.framework/Headers/fp.h:2145: error: 'rinttol' was not  
> declared in this scope
> /System/Library/Frameworks/CoreServices.framework/Frameworks/ 
> CarbonCore.framework/Headers/fp.h: In function 'long int roundtoll 
> (long double)':
> /System/Library/Frameworks/CoreServices.framework/Frameworks/ 
> CarbonCore.framework/Headers/fp.h:2183: error: 'roundtol' was not  
> declared in this scope
>
> This leaves me with questions:
> Is the Carbon framework POSIX C compliant?
> I suspect that it is not.  So how and where should _POSIX_C_SOURCE  
> be undefined?

I would suspect the right place is in the wx headers, probably in  
somewhere like setup.h. For the moment, could you write a hard-coded  
#undef  _POSIX_C_SOURCE in your setup.h (or setup.h.in) and see if it  
resolves the matter? If so, it'll just be a matter of writing a  
proper test before undefining it, either in configure.in or setup.h.in.

> If _POSIX_C_SOURCE is undefed does that bring back the strcasecmp  
> issue?

So long as the undefine is done before string.h gets included in wx,  
this should actually be a very simple solution to the problem.

Thanks,

Kevin

> Cheers
> Bill
>