[PATCH 0/2] [P/omap4] debian.ti-omap4/control* housekeeping

Tim Gardner tim.gardner at canonical.com
Thu Mar 22 20:13:09 UTC 2012


On 03/22/2012 01:16 PM, Ricardo Salveti wrote:
> On Thu, Mar 22, 2012 at 4:12 PM, Tim Gardner <rtg.canonical at gmail.com> wrote:
>> On 03/22/2012 06:11 AM, Paolo Pisati wrote:
>>>
>>> A fix and a cleanup for debian.ti-omap4/control.stub&c:
>>>
>>> 1) make linux-ti-omap4-headers provide linux-headers (thus fixing packages
>>> depending on it like dkms)
>>>
>>> 2) garbage collect unused entries
>>>
>>> BugLink: http://bugs.launchpad.net/bugs/960770
>>>
>>> Paolo Pisati (2):
>>>   UBUNTU: linux-ti-omap4-headers provides linux-headers too
>>>   UBUNTU: gargabe collect unused entries in control.stub.in
>>>
>>>  debian.ti-omap4/control.d/flavour-control.stub |    4 +-
>>>  debian.ti-omap4/control.stub.in                |   61
>>> +-----------------------
>>>  2 files changed, 3 insertions(+), 62 deletions(-)
>>>
>>
>> Ricardo - I think the correct fix for this is to add omap4 specific
>> recommends to dkms. Can you add this patch to dkms and give it a try ?
> 
> We could, but then the question is why this package should not provide
> linux-headers?
> 
> I know it's not coming from the official tree, but as this will happen
> with all the flavours we currently have (armada, ac100, linaro, etc).
> It might be good to fix it for all by just expecting the
> flavour-linux-headers to also provide linux-headers itself.
> 
> What do you say?
> 
> Thanks,

I don't think its the right solution. According to the Debian Policy
manual Provides: is meant to maintain the dependency graph when a newer
(and renamed) package provides the same support.

In this case, your powervr dkms package has a hard dependency on
linux-headers-omap4, and there is no other package that provides that
support.

I think the real issue are the Recommends: in the dkms package itself.
Given that it is now being used with a variety of kernels, the
recommendations are kind of bogus. Adding a new linux-image* and
linux-headers* recommendation for every new flavour of kernel doesn't
really scale. If dkms dropped all of its recommendations then I don't
think you'd be having the issue where apt-get wants to install
superfluous packages.

rtg
-- 
Tim Gardner tim.gardner at canonical.com




More information about the kernel-team mailing list