Questions about LTS backports to Hardy

[Colin Watson]

On Wed, Jun 24, 2009 at 01:27:01PM -0600, Tim Gardner wrote:
> Colin Watson wrote:
> > I think that makes sense, yes. You might want to be a bit careful about
> > the couple of binary packages that are -2.6.28 rather than
> > -2.6.28-ABI-FLAVOUR, but that are built by the main kernel source
> > package - IIRC linux-doc-2.6.28 and linux-source-2.6.28 fall into this
> > category. Just leaving those out of linux-meta-2.6.28 would be good
> > enough.
> 
> Your comment about binary packages is in reference to linux-meta,
> correct? I'll take some care to make sure there are no conflicts with
> existing Hardy meta packages.

I think we're in agreement but I'm not entirely sure. :-) Just to be
clear, what I meant was:

  linux (jaunty) produces binaries:
    linux-image-2.6.28-ABI-FLAVOUR
    linux-headers-2.6.28-ABI-FLAVOUR
    block-module-2.6.28-ABI-FLAVOUR-di
    [...]
    linux-doc-2.6.28
    linux-source-2.6.28

  linux-meta (jaunty) produces binaries:
    linux, linux-image
      (architectures with generic kernels)
    linux-FLAVOUR, linux-image-2.6.28-FLAVOUR
      (architectures without generic kernels, if you bother with those
      in this system)
    [...]
    linux-doc
    linux-source

  linux-2.6.28 (backported jaunty source package, whatever you call it)
  produces binaries with the same names as linux (jaunty):
    linux-image-2.6.28-ABI-FLAVOUR
    linux-headers-2.6.28-ABI-FLAVOUR
    block-module-2.6.28-ABI-FLAVOUR-di
    [...]
    linux-doc-2.6.28
    linux-source-2.6.28

  linux-meta-2.6.28 produces binaries with the same names as linux-meta
  (jaunty) BUT with "-2.6.28" added:
    linux-2.6.28, linux-image-2.6.28
      (architectures with generic kernels)
    linux-2.6.28-FLAVOUR, linux-image-2.6.28-FLAVOUR
      (architectures without generic kernels, if you bother with those
      in this system)
    [...]
    NOT linux-doc-2.6.28 because that would clash
    NOT linux-source-2.6.28 because that would clash

> Your comment did get me to thinking about the binary packages generated
> by the kernel. What do you suggest I do about linux-libc-dev, which will
> supersede an existing Hardy package? Shall I rename it, or just not
> produce it? I think it ought to be superfluous.

I'd ditch it - we don't want it used for userspace builds, and I don't
think it'll be all that interesting for users for the most part. If you
do find that you have demand for it then linux-libc-dev-2.6.28 would be
an OK name, but the packaging could be a bit delicate - it'd have to
conflict with linux-libc-dev, and it wouldn't surprise me if glibc had
to be rebuilt against it. Probably best avoided.

-- 
Colin Watson                                       [cjwatson at ubuntu.com]