On 05.05.2011 13:34, Phil Blundell wrote:
> On Thu, 2011-05-05 at 13:21 +0200, Simon Busch wrote:
>> On 05.05.2011 11:50, Phil Blundell wrote:
>>> On Wed, 2011-05-04 at 19:00 +0200, Simon Busch wrote:
>>>> I alreay searched through the cmake documentation and it says that the
>>>> rpath is set with the content of the CMAKE_INSTALL_RPATH variable. The
>>>> default value of CMAKE_INSTALL_RPATH is ""[1]. As currently nothing sets
>>>> it to another value the rpath is never set during the build for the
>>>> resulting binaries. I send another patch which does the solution
>>>> mentioned above the right way.
>>>
>>> Oh right, I see. So cmake is actually stripping the rpaths during the
>>> install step unless you tell it that you want them included?
>>>
>>> In that case, yeah, your plan seems reasonable.
>>
>> Ok, so you are fine with the patch?
>
> Yup. My only remaining concern was that it seemed a bit unwholesome to
> do:
>
> + echo "set( CMAKE_INSTALL_RPATH ${OECMAKE_RPATH} )" >>
> ${WORKDIR}/toolchain.cmake
>
> when OECMAKE_RPATH might be empty, and maybe it'd have been better to
> bracket that echo with "if [ -n ${OECMAKE_RPATH} ] ...". But if you've
> tested it and cmake does the right thing then I guess it's fine as it
> stands.
>
> Acked-by: Phil Blundell <[email protected]>
Ok, thanks for your help. Patch is now pushed but somehow cgit got the
line breaks wrong while local git is showing them correctly ...
regards,
Simon
_______________________________________________
Openembedded-devel mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
