Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: x86 GENERIC kernels unlobotomized

On Tue, Aug 09, 2011 at 07:18:59PM +0200, Marc Balmer wrote:
> (...)
> It's a bit off-topic, but I find the current state of affairs, having an
> open directory with the modules, the right way.  So I can easily add a
> module to the system module area, without needing to create a tgz file.

I wasn't thinking of needing to create a tgz file. All *I* would require
to stop shooting myself in the foot every now and then would be release to create a single "kernel" set, that includes everything
that is necessary to go ahead and reboot with it. You do NOT want to install
the base set with the old kernel (your binaries might end up requiring
stuff your current kernel cannot deliver), so putting the modules in the
"base" set seems inadequate at best. I try remembering copying the kernel
and then doing a tar -xzf on base to only extract the modules, but it's more
cumbersome than it should be (to install from a binary release) where the
modules could simply be part of the kernel set instead of being part of
the base set (they are NOT base system after all!).

Again, the suggestion, AFAICT, never included forcing you to install from
binary release sets instead of installing from source, distribution results,
pax, -E -D / or whatever. All "we" (me, too, IIRC) were asking for
was that the binary blob that makes up the kernel, whether it is spread over
multiple files or a single one (i.e., MODULAR vs. MONOLITHIC), end up packed
up nicely together in the kernel set.

> I think we have more prevalent problems to solve with modules than where
> to put the modules and in which form.  We need to do the work and
> modularize what is not yet, e.g.

Alright, I cannot help with the more prevalent problems. How about I try
moving the modules to the kern sets, would such a patch be accepted?


Home | Main Index | Thread Index | Old Index