ACK/Cmnt: [SRU][f/linux-meta-*][PATCH 0/6] Fix missing kernel-testing--*--modules-extra--preferred

Kleber Souza kleber.souza at canonical.com
Fri Jun 11 09:46:07 UTC 2021


On 11.06.21 11:23, Andrea Righi wrote:
> On Thu, Jun 10, 2021 at 06:00:48PM +0200, Kleber Sacilotto de Souza wrote:
>> BugLink: http://bugs.launchpad.net/bugs/1931576
>>
>> [Impact]
>> ADT sets up the test environment by installing the image and modules-extra
>> package via the virtual provides
>> 'kernel-testing--linux-<derivative>--full--preferred' and
>> 'kernel-testing--linux-<derivative>--modules-extra--preferred'. If those are not
>> available, it falls back to using the real meta names as
>> 'linux-image-<derivative>' and 'linux-modules-extra-<derivative>'.
>>
>> The 5.4 cloud kernels in Focal are not providing the virtual packages for
>> 'kernel-testing--linux-<derivative>--modules-extra--preferred'. When the tests
>> falls back to 'linux-modules-extra-<derivative>', the new 5.8 cloud kernels are
>> pulled for the tests as they are now the default sources.
>>
>> The oracle kernels in focal don't need fixing as they do not produce a
>> linux-modules-extra-* package.
>>
>> ADT log:
>> https://autopkgtest.ubuntu.com/results/autopkgtest-focal/focal/amd64/a/acpi-call/20210609_092540_0a7a8@/log.gz
>>
>> [Fix]
>> The meta packages need to be fixed to provide the necessary virtual packages for
>> testing.
>>
>> Although the 5.8 kernels don't need this fix, I'm including it for
>> consistency.
>>
>> [Where problems could occur]
>> Adding new virtual meta packages could clash with existing names provided by
>> other sources. However, if they existed ADT would have pulled them.
>>
>> Kleber Sacilotto de Souza (1):
>>    UBUNTU: [Packaging] Fix provides for *--modules-extra--preferred
>>
>>   debian/control.d/azure | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> Makes sense to me. How about 5.11 and the next kernels? Do we need to
> make sure this change is applied also to them for consistency?

Thanks Andrea.

Yeah, we should apply to the newer kernels as well even if it's not
strictly necessary now. Another advantage is that carrying this forward
would make them produce the correct virtuals when the next LTS release
comes.

I'll prepare fixes for the 5.11 trees and send in a separate thread.

Kleber



More information about the kernel-team mailing list