Bug#582776: initramfs-tools: pointless generation of an initrd for initrd-free custom kernels
On Tue, Jun 08, 2010 at 03:35:36PM +0200, Michael Prokop wrote:
> * Darren Salt <ds@youmustbejoking.demon.co.uk> [Son Mai 23, 2010 at 03:26:28 +0100]:
>
> > While installing a locally-built kernel (which doesn't require an initrd), I
> > found that /etc/kernel/postinst.d/initramfs-tools was nevertheless generating
> > an initrd for that kernel.
>
> > This causes breakage here. I'm using lilo and a slightly-modified
> > version of the example symlink_hook script provided by kernel-package; if an
> > initrd is present, the symlink name is different. (I'm using the symlinks so
> > that I don't have to edit lilo.conf.)
>
> > (I also have a stock kernel installed for occasional testing purposes.
> > Mostly, I just use my custom kernels.)
>
> > I find this sufficient to handle packages generated by kernel-package:
>
> > --- /etc/kernel/postinst.d/initramfs-tools~ 2010-04-08 04:53:09.000000000 +0100
> > +++ /etc/kernel/postinst.d/initramfs-tools 2010-05-23 15:06:44.114662432 +0100
> > @@ -11,6 +11,10 @@
> > if [ -n "${KERNEL_PACKAGE_VERSION}" ]; then
> > bootdir=$(dirname "$2")
> > bootopt="-b ${bootdir}"
> > + # check whether an initrd is needed
> > + if ! grep -q '^CONFIG_BLK_DEV_INITRD=y$' /boot/config-"$1"; then
> > + exit 0
> > + fi
> > else
> > # official Debian linux-images take care themself
> > exit 0
>
> Looks fine to me.
>
> maks, any objections against adding this patch? I'd take care of it.
yes wrong.
don't assume to look on /boot/config-version for respective image.
didn't see better idea here so close.
Reply to: