[klibc] Re: klibc and 2.6

[Bryan O'Sullivan]

On Sun, 2003-05-25 at 17:07, Kai Germaschewski wrote:

> I basically did a complete integration of klibc into kbuild two or so 
> months ago - however, I didn't touch it since, so it won't apply cleanly. 
> I can dig out the bk tree, if you're interested.

I've put my own latest integrate-klibc-into-kbuild BK patch up at

    http://www.speakeasy.org/~bos/files/klibc-2.5.69-2.patch.bz2

This integrates most of CVS head klibc into kbuild for 2.5.69, but it
has some major "features", some of which I need help with.

klibc itself and all of its auxiliary bits (ipconfig, nfsmount, and
kinit) now build.

However, the way in which they build is very obviously not what Greg
intended when he put the original patches together some months ago.  I
don't know enough about kbuild to fix this.  The current state of the
Makefiles corresponds to the minimal grubbing about I could do to get
things to work at all.

All of klibc is always rebuilt, and I don't know why.  The dependencies
are not getting set or tracked properly by usr/lib/Makefile or one of
the new Makefile.* in scripts/, but I do not know how to debug this
problem.

This patch breaks the regular kernel build in a rather obvious manner. 
It issues complaints like this:

        mv: `arch/i386/kernel/.tmp_process.o' and `arch/i386/kernel/.tmp_process.o' are
        the same file

for every .o file built.  Once again, I don't know what is wrong here.

Kai, I think I am about 80% of the way to having the entire thing sane,
but I am quite lost right now; it will take me a lot of work to figure
out how to unbreak what's left.  If you can take a look at the patch and
give me a few pointers for things I can fix, I'd be very grateful.

	<b