Re: more on kde4 build
[email protected] Sun, 13 Nov 2005 18:51:11 -0500
| Newsgroups | gmane.comp.kde.darwin |
|---|---|
| Message-ID | <[email protected]> |
On Wednesday 09 November 2005 09:25 pm, you wrote:
> > For the sake of trying to generate discussion i was able to compile
> > kdecore on panther with some tiny modifications. For some reason bksys
> > didn't detect sys/mman.h nor net/if.h, after some hack-ish modifications
> > kdecore builds.
>
> Hmm... can't speak for sys/mman.h but I specifically wrote a test for
> it that should work. =A0What do you get in config.log when it's trying
> to find them?
i finally had time to look at why scons was not finding sys/mman.h. When sc=
ons=20
checks for sys/mman.h it only compiles this:
#include<sys/mman.h>
However there are macros in mman.h that require types.h, at least that's th=
e=20
case on panther; so the compile fails and HAVE_SYS_MMAN_H is not defined.=20
This is the same problem with net/if.h. So i added sys/types.h to the array=
=20
passed to conf.CheckHeader( ... )
113 content +=3D define_line(header_define,
114 conf.CheckHeader(["sys/types.h",header]))
=2E...
119
120 content +=3D define_line("HAVE_NET_IF_H",=20
conf.CheckHeader(['sys/types.h','sys/socket.h','net/if.h']))
This seems to do the trick on panther. It may seem a little overkill to hav=
e=20
sys/types.h included for every header file check, but i don't think there a=
re=20
any bad sideeffects.=20
=2Dbkn