[OE-core] GCC fails when SDK is not extracted to /usr/local

[Khem Raj]

On Thu, Dec 8, 2011 at 11:11 AM, Mark Hatle <mark.hatle at windriver.com> wrote:
> On 12/8/11 1:04 PM, Tasslehoff Kjappfot wrote:
>>
>>
>> On Dec 8, 2011, at 3:56 PM, Tom Rini wrote:
>>
>>> On Wed, Dec 7, 2011 at 12:16 AM, Tasslehoff Kjappfot
>>> <tasskjapp at gmail.com>  wrote:
>>>>
>>>> gcc fails when I extract my SDK to another place than /usr/local.
>>>>
>>>> Output when it fails:
>>>>
>>>> $ ./arm-angstrom-linux-gnueabi-gcc
>>>> bash: ./arm-angstrom-linux-gnueabi-gcc: No such file or directory
>>>>
>>>> Readelf output:
>>>>
>>>> $ readelf -d arm-angstrom-linux-gnueabi-gcc
>>>>
>>>> Dynamic section at offset 0x34394 contains 21 entries:
>>>>  Tag        Type                         Name/Value
>>>>  0x00000001 (NEEDED)                     Shared library: [libc.so.6]
>>>>  0x0000000f (RPATH)                      Library rpath:
>>>>
>>>> [/usr/local/angstrom-eglibc-i686-armv7a/sysroots/i686-angstromsdk-linux/usr/lib/armv7a-angstrom-linux-gnueabi:/usr/local/angstrom-eglibc-i686-armv7a/sysroots/i686-angstromsdk-linux/lib]
>>>>
>>>> Is it the RPATH here that's making trouble for me?
>>>
>>>
>>> Not exactly.  And RPATH is needed to find some of the shared libraries
>>> that are needed.  I thought we had been running chrpath over these to
>>> make them relocatable, however.
>>>
>>> --
>>> Tom
>>>
>>
>> Does this RPATH indicate that its not relocatable? Anything else I can
>> check?
>
>
> It should be using RUN_PATH  $ORIGIN to relocate itself based on the initial
> run-time location.
>
> It sounds like something is not being run as it should for the SDK
> component(s).

it should come by inheriting relocatable class I think and this class is being
inherited in cross|native|nativesdk classes.