[quantal PATCH 0/1] Beef up linux-image-extras-* postinst/postrm
Tim Gardner
tim.gardner at canonical.com
Tue May 29 12:29:42 UTC 2012
On 05/29/2012 02:37 AM, Andy Whitcroft wrote:
> On Mon, May 28, 2012 at 12:40:33PM -0600, Tim Gardner wrote:
>> On 05/28/2012 12:13 PM, Andy Whitcroft wrote:
>>> When we install linux-image-generic-* and linux-image-extras-generic-*
>>> we end up installing linux-image-generic-* first due to internal
>>> dependancies, we then regenerate the initramfs, we then install
>>> linux-image-extras-generic-*, indicate we would like to rebuild the
>>> initramfs again but do not (it is elided) and complete. This potentially
>>> leaves us with a half formed initramfs.
>>>
>>> After discussions it is clear we should actually be doing the full post
>>> kernel install hook runs. Following this email is a patch to do this.
>>>
>>> It should be noted that for a split kernel, ie all main kernels in
>>> Quantal we will build the initramfs twice. This is at least correct if
>>> sub-optimal. The correct solution is to perameterise the update-initramfs
>>> hook so that the kernel can delay the rebuilds to then end of the apt run;
>>> very likely update-grub would also have to become a hook. I will discuss
>>> this further with foundations.
>>>
>>> Proposing for Quantal.
>>>
>>> -apw
>>>
>>> Andy Whitcroft (1):
>>> UBUNTU: [Config] linux-image-extras needs full postinst
>>>
>>> debian/control-scripts/postinst.extra | 154 +++++++++++++++++++++++++++++++--
>>> debian/control-scripts/postrm.extra | 143 ++++++++++++++++++++++++++++--
>>> debian/rules.d/2-binary-arch.mk | 5 +-
>>> 3 files changed, 284 insertions(+), 18 deletions(-)
>>>
>>
>> Why can't the extras package just run
>> /var/lib/dpkg/info/linux-image-<ABI>-<flavour>.{postinst,postrm} directly ?
>
> Those postinst etc do the links as well, and IIRC they use the current
> package name, I could parameterise the links support in the existing
> postinst and simply share them.
>
> -apw
I think the less extra code (especially perl) we carry, the better.
--
Tim Gardner tim.gardner at canonical.com
More information about the kernel-team
mailing list