Chosing the C and C++ compilers

Akim Demaille akim at lrde.epita.fr
Thu Nov 6 07:17:58 PST 2008


Le 6 nov. 08 à 16:05, Anders F Björklund a écrit :

> Akim Demaille wrote:
>
>> Ah, OK, didn't think about that, thanks!  Then in that case I  
>> suppose that there are means to tell /usr/bin/gcc-4.0 what arch I'm  
>> aiming for (686, ppc, or universal).  I suppose via something like - 
>> march.  In which case, it's even better than using the fully  
>> qualified compiler: /usr/bin/gcc-4.0 with the appropriate flags  
>> allows to build for the three cases.
>>
>> But apparently in the default mode it compiles for the build  
>> machine, which is not what I need.  So I guess the question is  
>> rather: do I have a means to tell /usr/bin/gcc-4.0 via  
>> macports.conf the set of archs I want to compile for?
>
> -march and -mtune set which machine you want to aim the  
> optimizations for
> -arch and -m32/-m64 set which architectural platform you want to  
> compile for
>
> And when you are done instructing GCC, you also need to tell the other
> parts of the build system what the platform is - for instance  
> configure...
>
>
> There are some hacks in configure and universal for setting these,
> but they don't have really a simple configuration for cross-compiling.
>
> The default MacPorts setup is to compile everything locally only,
> so building for other machines is something of an after-thought...

Thank you, but there seems to be some misunderstanding here: I am not  
asking for help on how I should talk to GCC, but rather how I should  
tell macports to use the compiler in a more useful way.  I do think  
that rather than leaving gcc-4.0 what's to be done, macports should  
use its knowledge about the variants asked by the user to decide what  
flags to pass to the compiler, in which case distcc comes for free.

That would be the ideal.

Meanwhile, I'm asking how I can educate macports to stop using a hard  
coded path for gcc and let *me* do what I know I want to do.



More information about the macports-users mailing list