Re: Grub vs. linux-image-2.6.32 conundrum
On Wed, Mar 31 2010, Stephen Powell wrote:
> On Wed, 31 Mar 2010 11:08:29 -0400 (EDT), John Hasler wrote:
>> Stephen Powell wrote:
>>> If there is a bug...
>>
>> There clearly is.
>>>
>>> But as for it's operation, it is working as designed.
>>
>> Design errors are still bugs.
>
> The main difference between a bug and a feature is that a feature is
> documented and a bug is not. So perhaps you are right. I can find
> no official documentation for /etc/kernel-img.conf as used by the
> maintainer scripts which ship with official Debian stock kernel image packages.
> There is some documentation for the version of /etc/kernel-img.conf
> which is used by the maintainer scripts which are packaged with kernel
> image packages created by make-kpkg in the kernel-package package,
> but that clearly doesn't apply here.
>
> As best as I can tell, kernel-package was at one time used by the
> Debian kernel team to create official Debian stock kernel image
> packages. But at some point in the past there was a parting of the
> ways, and the Debian kernel team started using other tools to create
> official Debian stock kernel image packages.
Arguably, at this point, they should have also stopped using
/etc/kernel-img.conf (perhaps still parsing it as a fallbacK), and
started using and documenting a _new_ file. If that had been done,
with the postinst only reading /etc/kernel=img.conf when the new config
file was not present, would have allowed for a graceful transition to
the new, differently documented, configuration file.
> What I learned about /etc/kernel-img.conf I learned from reading the
> man page that comes with the *Lenny* version of kernel-package.
> However, starting with the Squeeze version of kernel-package, there is
> a major philosophical departure from the past. The new philosophy of
> the maintainer scripts that are packaged with a kernel image package
> created by make-kpkg is that *no* post-installation tasks such as
> creating an initial RAM filesystem, updating the symlinks, or
> re-running the boot loader will be performed. If you want those
> things, you need to do them in a hook script. The maintainer scripts
> that ship with stock kernel image packages still support most of these
> options. Documentation for most of these options has been removed
> from the man page that ships with the Squeeze version of
> kernel-package. The closest thing to documentation for the Squeeze
> version of /etc/kernel-img.conf, as used by the maintainer scripts for
> official Debian stock kernel image packages, is the man page for
> kernel-img.conf that ships with the *Lenny* version of kernel-package.
>
> This is not a good situation, and it should be addressed. The problem
> is, against what package would you open a bug report, since the file
> does not belong to a package? The file is referenced by the
> maintainer scripts of *every* stock kernel image package for *every*
> architecture, as well as by some other packages, such as the
> update-initramfs script of initramfs-tools. (By the way, the fact
> that "do_bootloader = yes" is *not* honored for initial RAM filesystem
> *creation*, but *is* honored by an initial RAM filesystem *update*,
> may be a bug in the update-initramfs script of the initramfs-tools
> package.)
manoj
--
"The algorithm to do that is extremely nasty. You might want to mug
someone with it." -- M. Devine, Computer Science 340
Manoj Srivastava <srivasta@acm.org> <http://www.golden-gryphon.com/>
4096R/C5779A1C E37E 5EC5 2A01 DA25 AD20 05B6 CF48 9438 C577 9A1C
Reply to: