Re: Kernel Source Tree Size: Platform Subsets?

Keith Rohrer (
Fri, 27 Feb 1998 11:08:20 -0600 (CST)

And lo, Jim Dennis saith unto me:
> Is our source tree getting large enough to consider some option
> to split the full tree into "architecture subsets." If I just want
> the x86 code would it be neat to be able to say:
> make archsubset x86
> ... to produce a tarball of just the x86 code?
(1) We must not make the amount of data to be archived much larger.
(2) We must not make patching any more painful at all.
(3) We must split off enough stuff to justify the number of pieces.

If we can do that, the developers who need the whole thing can
learn to use mget rather than get. I just don't think we can
manage 2 and 3 by splitting off the architecture directories.

> (In this case we might even have some subsets of the x86 -- such
> as GGI vs. non-GGI).
Ick. This would turn into two separate versions real fast.

> How much trouble would it be to do that? (I guess you'd have to
> have a set of "meta makefiles" that would specify which portions
> of the source tree were needed by which source subsets).
The only practical objections I remember to splitting off e.g. the
CD-ROM drivers were the not-having of the split-off pieces by
the developers (mget is your friend...) and the difficulty of
applying whole-tree patches to a partial tree (this would take at
least a smarter patch tool and probably a new patch format as well).
There also may be valid objections I don't remember or which are


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to