Re: [gentoo-dev] kernel-mod.eclass addition + clean up

[Henrik Brix Andersen]

Hi,

On Sun, 2004-10-31 at 21:41, Daniel Drake wrote:
> (not all of these points are related to your changes, but it would be nice to
> get them ironed out while we are working on it)

Thanks for providing feedback. I agree that we might as well solve all
known issues while we're at it.

> I'm not too sure about kernel-mod_src_compile. There isn't a generic argument
> (e.g. KERNEL_DIR) which makefiles tend to follow when specifying the kernel
> source. If anything we should go for a hackish way at trying to support as
> many builds as possible, e.g. setting KDIR, KERNDIR, KERNEL_DIR, KERNEL,
> KERNEL_SRC, .., to /usr/src/linux and O, M, OUTPUT, SYSOUT, to ${S} .. but
> that may well break some builds. I think we should leave this out altogether.

Yeah, I was uncertain about kernel-mod_src_compile() as well. Just
didn't want to take it out since sys-apps/cloop depends on it. We could
leave it out and fix sys-apps/cloop accordingly?

> kernel-mod_subdir_to_m requires some extra thought, its not always a simple
> search and replace. (e.g. autotooled Makefiles often use an internal SUBDIRS
> variable if I remember right). In most cases this would be useful, but there
> are always some oddities.

Yes, that's why I included kernel-mod_need_subdir_to_m(). Ebuilds which
require more than the simple sed statement can do:

	if kernel-mod_needs_subdir_to_m
	then
		...
	fi

Will this do?

> kernel-info_getversion - is this needed? portage.py provides the variable $KV
> which should give the same info. I think we should only keep one or the
> other. If we keep kernel-info_getversion then it needs to be extended to read
> 2.6.9's new LOCALVERSION stuff (the internal portage version has been updated
> in CVS).

Many ebuilds rely on having more than ${KV} available.
kernel-info_getversion() provides ${KV_MAJOR}, ${KV_MINOR} and
${KV_PATCH}.

But you're right: kernel-mod_getversion() needs to be updated to
understand LOCALVERSION.

> kernel-info_configoption* - yes, I know I wrote these, but I think we should
> change the .config checking a bit:
> 	[ -e "${KERNEL_DIR}/.config" ] || die "Kernel has not been configured yet"
> Instead of the above, we should perhaps create another function and use it to
> print out a more informative error message if no .config file has been found.

Good idea. I'll write a function for that.

> Other than that, I like the proposed split. Thanks!

Thanks again for providing feedback. I'll write a new version of the two
eclasses implementing the above mentioned changes tomorrow.

Stay tuned,
Brix
-- 
Henrik Brix Andersen <brix@g.o>
Gentoo Linux