[oe] MACHINE_KERNEL_PR issue

[Frans Meulenbroeks]

2009/10/26 Frans Meulenbroeks <fransmeulenbroeks at gmail.com>:
> 2009/10/26 Frans Meulenbroeks <fransmeulenbroeks at gmail.com>:
>> 2009/10/26 Koen Kooi <k.kooi at student.utwente.nl>:
>>> On 26-10-09 16:37, Frans Meulenbroeks wrote:
>>>>
>>>> Hi,
>>>>
>>>> While creating a recipe which generates a module I bumped into the
>>>> following issue.
>>>> If I do
>>>> inherit module
>>>> the package name uses MACHINE_KERNEL_PR and not PR.
>>>> I understand the reason for this: if the kernel is modified you want
>>>> the proper modules.
>>>> However in a recipe this means that if the recipe changes for whatever
>>>> reason that technically one should increase all MACHINE_KERNEL_PR
>>>> values (as otherwise the new version is not used).
>>>> In my eyes this is not very desirable.
>>>>
>>>> Any opinions how we should fix this?
>>>
>>> koen at dominion:/OE/org.openembedded.dev$ grep PR_append recipes/ti/*
>>> recipes/ti/ti-cmem-module_2.24.01.bb:MACHINE_KERNEL_PR_append = "a"
>>> recipes/ti/ti-dm355mm-module_1.13.bb:MACHINE_KERNEL_PR_append = "a"
>>> recipes/ti/ti-dmai_svn.bb:MACHINE_KERNEL_PR_append = "c"
>>> recipes/ti/ti-dsplink-module_1.61.3.bb:MACHINE_KERNEL_PR_append = "a"
>>> recipes/ti/ti-lpm-module_2.24.01.bb:MACHINE_KERNEL_PR_append = "a"
>>> recipes/ti/ti-sdma-module_2.24.01.bb:MACHINE_KERNEL_PR_append = "a"
>>
>> Ah, ok didn't know about that one and when coined this on #oe no-one
>> mentioned this solution.
>> Thanks alot! Frans
>>
>
> Two more things:
>
> We have about 83 recipes that do inherit module.
> and 12 that have MACHINE_KERNEL_PR (10 ti ones, linux omap pm and
> open-iscsi which I just added)
> so ti seems 71 have a problem
>
> and why not change the class and add _$(PR) or so instead of _append ?
As I saw no replies on this, replying to it myself to bring it on the
table again.

Can we agree on a policy wrt versioning of kernel modules.
My preference is a solution which uses ${MACHINE_KERNEL_PR} '+' ${PR}
where the recipe provides PR (just like for non-kernel recipes).
Can we agree on this (and can someone implement this, or do you trust
my measly python skills to do this myself :-) )

Frans