On Sunday 03 April 2005 13:21, Andrew de Quincey wrote: > On Sunday 03 April 2005 13:18, Andrew de Quincey wrote: > > On Sunday 03 April 2005 11:35, Stuart Longland wrote: > > > Andrew de Quincey wrote: > > > > On Sunday 03 April 2005 03:10, Andrew Gaffney wrote: > > > >>Andrew de Quincey wrote: > > > >>>Hi, I hope this is the correct place to send this. > > > >>> > > > >>>I have setup distcc in my network. Most of the hosts are i686, but > > > >>> one of them is x86_64. I wish to distribute tasks to the i686 > > > >>> machines from the x86_64 machine. This involves installing a cross > > > >>> compiler on the i686 machines for x86_64 tasks. Crossdev is > > > >>> overkill for this - I don't actually _need_ glibc and the like on > > > >>> the client machines for a simple distcc environment. It really adds > > > >>> to the upgrade time when emerge sync; emerge worlding. > > > >> > > > >>crossdev -s1 -t <x86_64 CHOST> > > > > > > > > According to the docs, that doesn't compile a c++ compiler: > > > > > > > > -s1, --stage1 Build a C compiler (no libc/C++) > > > > > > And guess what... no libc... no c++.
The dependency on libc is not in the c++ compiler. This is what led me to realise I could do this. The dependency on libc comes from when it builds the target libraries - libgcc & libstdc++ and the like. First of all, gcc builds all the compilers - C and C++. After its done that, it compiles libgcc etc - and as I said earlier, this is what needs a libc for the target machine. My patch simply disables the compilation/installation of the target machine's libraries. All you end up with are the binaries of the cross compilers; which is all you really want for distcc. Have you had a chance to try it out yet? Lemme know if you like it, and I'll see if I can clean it up more. -- gentoo-dev@gentoo.org mailing list