[meta-virtualization] [PATCH] Revert "runc: Don't override GOARCH or GOROOT"

[Paul Barker]

On Tue, Oct 10, 2017 at 12:30 AM, Paul Barker <pbarker at toganlabs.com> wrote:
> On Tue, Oct 10, 2017 at 12:02 AM, Bruce Ashfield
> <bruce.ashfield at gmail.com> wrote:
>> On Mon, Oct 9, 2017 at 5:56 PM, Paul Barker <pbarker at toganlabs.com> wrote:
>>>
>>> It seems like we have conflicting changes :( Reverting this patch will
>>> break cross-compilation of runc-opencontainers for x86-64 with musl
>>> libc again.
>>>
>>> I'll try cross-compiling both runc variants for arm, aarch64, x86 and
>>> x86-64 and see if I can find a solution that works for everything.
>>>
>>
>> Agreed.
>>
>> I'll hold off on any revert until I hear from your tests.
>>
>> For the configs I'm using everything is building with the current set
>> of changes.
>>
>> Bruce
>>
>
> I can re-produce the failure by building runc-docker for
> raspberrypi3-64. However, runc-opencontainers builds successfully for
> raspberrypi3-64. So at worst we can add a do_compile_prepend() in the
> recipe for runc-docker (but not the common .inc file). However, the
> whole thing looks wrong, GOROOT shouldn't be pointing at the native
> sysroot at all. There's probably something strange going on in the
> runc-docker makefile.
>

I've found that the issue is with building the recvtty demo/reference
application within runc-docker. That's under a contrib directory in
the source tree and isn't a core part of the runc application. If I
patch out the line in the Makefile which builds recvtty then the build
succeeds for all platforms I build tested (qemux86, qemux86-64,
raspberrypi, raspberrypi2, raspberrypi3, raspberrypi3-64).

So I'm going to propose we just patch out the compilation of recvtty
for this recipe unless someone is actually using it. Aníbal, is this
ok for you?

Thanks,

-- 
Paul Barker
Togán Labs Ltd