Yes, the bbappend is breaking the packaging so should fix its own mess.

Ross

On 4 July 2017 at 08:36, Gordan Markuš <[email protected]> wrote:

> On 2017-07-03 15:56, Burton, Ross wrote:
>
>
> On 3 July 2017 at 13:34, Gordan Markuš <[email protected]>
>  wrote:
>
>> -FILES_${PN}-dev += "${libdir}/cairo/*.so"
>> +FILES_${PN}-dev += "${libdir}/cairo ${libdir}/cairo/*.so"
>>
>
> I just spent five minutes working out *why* this fixes the QA issue you
> were getting.
>
> A better fix would be to delete ${libdir}/cairo using:
>
>   rmdir ${libdir}/cairo || :
>
> or
>
>   rmdir --ignore-fail-on-non-empty ${libdir}/cairo
>
> (former is POSIX compliant, but I can't think of a situation where that is
> only a hypothetical problem)
>
> Then in the commit message you can say that the directory is always
> created but it only has content if you disable cairo-trace.  At least, I
> hope you are explicitly disabling cairo-trace as there's no other way I can
> see for you go get this.
>
> Ross
>
> Thanks for the suggestion Ross, and you are right this only happens if
> there is a bbappend that removes the 'cairo-perf-utils' package. An example
> of this is the meta-ivi layer.
> Experimenting a bit it seems to me that even by adding the directory
> removal, in case it is empty, the do_package step still fails. By
> inspecting further I can observe that the directory is in fact not being
> removed but left empty which is quite puzzling.
> After doing some thinking I believe it might be better to add the removal
> of the directory and the package files in the bbappend, do you agree with
> this? Here is the PR to meta-ivi that does that https://github.com/GENIVI/
> meta-ivi/pull/45.
>
-- 
_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.openembedded.org/mailman/listinfo/openembedded-core

Reply via email to