Joe Ciccone:
> Roman I Khimov wrote:
> > I think there is a misunderstanding here. :) What I see currently in
> > CLFS and CBLFS is that we cross-compile some basic temporary system
> > just to boot/chroot into it and then compile natively everything on
> > target. Second part doesn't differ that much from LFS/BLFS. What I want
> > to see is full cross-compilation, every package builds on some build
> > machine for target machine and the result is not something temporary
> > but final system.
>
> That would be very cool. I managed to cross-compile openssh (openssl)
> and through all the dependencies for gtk+-2 when I got stuck. I suppose
> the next time I do that I'll document it. Hopefully, it will get big
> enough for a book. I don't think a wiki, like cblfs, would be
> appropriate for it. Since there are so many critical parts to a build
> like this. A lot of the times you need to compile a package twice to
> build binaries that are run during the build.

Yeah, I sure know this, because I did such compile for one project. It 
should be noted though, that it was was based on Fedora Core 5 packages and 
used rpmbuild for building. Still that required lots of patches plus some 
more utilities for cross-toolchain. One of the most interesting things is 
build machine's pkg-config that doesn't work at all for cross-compilation 
without patching.

My build used some ugly hacks, because I've just had no time to fix 
everything properly, that's why I'm very interested in a book that would 
show how to "do the right thing" (c). I can even help in writing one, 
although I'm very busy currently with other stuff.

And I remember when I was starting that project I've hit CLFS and have asked 
myself a question - "why only temporary system is cross-compiled?". 
Finished that I knew why, because it's hell. Even there is such a good 
thing like autobuild, it looks like only GNU folks know how to use it 
right. :)

> I think we could integrate this kind of information into cblfs, if we
> wanted to, but it would probably just clutter up the pages more then
> they'd need to be.

To me that is just what CLFS and BCLFS (or CBLFS, doesn't matter) should be. 
Yeah, it's hard, but it's a real cross-environment from scratch.

-- 
                            Roman
 http://roman.khimov.ru                                              .o.
mailto: [EMAIL PROTECTED]                                              ..o
gpg --recv-keys 0xE5E055C3 --keyserver hkp://subkeys.pgp.net         ooo

Attachment: pgpXe5no5KR0y.pgp
Description: PGP signature

_______________________________________________
Clfs-dev mailing list
[email protected]
http://lists.cross-lfs.org/cgi-bin/mailman/listinfo/clfs-dev

Reply via email to