Dne 25.6.2013 22:37, Björn Persson napsal(a):
Petr Pisar wrote:
> On 2013-06-25, Vít Ondruch <vondruch(a)redhat.com> wrote:
>> Is there some common practice, where to place architecture specific
>> header files? From output of the following command, I can't see any
>> such place.
>>
> I dont't think so. SDL moves architecture specific header file into
> a private directory:
>
> # Rename SDL_config.h to SDL_config-<arch>.h to avoid file conflicts
> on # multilib systems and install SDL_config.h wrapper
> mv %{buildroot}/%{_includedir}/SDL/SDL_config.h
> %{buildroot}/%{_includedir}/SDL/SDL_config-%{_arch}.h
> install -m644 %{SOURCE1} %{buildroot}/%{_includedir}/SDL/SDL_config.h
>
> And installs a wrapper (SOURCE1) to replace the original location. The
> wrapper is just a bunch of #if branches:
>
> #if defined(__i386__)
> #include "SDL_config-i386.h"
> #elif defined(__ia64__)
> #include "SDL_config-ia64.h"
> [...]
> #endif
As another example, Glib puts an architecture-specific header under
libdir and expects you to use pkg-config to make the compiler find it:
Using external tools for configuration, that is what I want to avoid. Of
course for Ruby, I could use mkmf, which will create suitable Makefile
for me, or pkgconfig. But why should I create complex makefile or
configuration script for such simple program I posted in my initial email.
Vít