[OE-core] [PATCH] kernel-devsrc: Don't package kernel meta-data

[Bruce Ashfield]

On Thu, Aug 4, 2016 at 3:48 AM, Jacob Kroon <jacob.kroon at gmail.com> wrote:

> On 08/04/2016 03:41 AM, Bruce Ashfield wrote:
>
>> On Wed, Aug 3, 2016 at 5:33 AM, Jacob Kroon <jacob.kroon at gmail.com
>> <mailto:jacob.kroon at gmail.com>> wrote:
>>
> [...]
>
>>     Ignore this patch, I'm suspecting that .kernel-meta/ directory
>>     shouldn't exist in ${STAGING_KERNEL_DIR} to begin with ?
>>
>>
>> It doesn't hurt anything by being there, but you can either intercept it
>> at the
>> time we stage the common kernel parts, or when you package devsrc ..
>> either
>> will work.
>>
>
> Since commit 3b49416fc7a7ee9bfe722f2e6089aa18df41dc58, linux-yocto.inc
> does:
>
> do_install_append(){
>         if [ -n "${KMETA}" ]; then
>                 rm -rf ${STAGING_KERNEL_DIR}/${KMETA}
>         fi
> }
>
> But I think this accidentally got broken by a later commit
> 3b2b4eef0fd06099342fcc2dfa35ecf37840c090, which changed the directory
> name by conditionally prepending a ".":
>
> kmeta = d.getVar( "KMETA", True ) or "meta"
> if not os.path.exists(kmeta):
>     kmeta = "." + kmeta
>
> So it seems to me that the right thing to do here would be to fix and
> move the do_install_append() to kernel-yocto.bbclass. Bruce, what do
> you think ?
>

Yep, that rm -rf could be both ${KMETA} and .${KMETA}. If either exist,
they should be removed. Whether it is in the .inc or the bbclass isn't
significant. I'd leave it where it is as smallest footprint action.

Bruce


>
> Regards Jacob
>



-- 
"Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end"
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20160804/60d457c0/attachment-0002.html>