Bug#835170: transition: protobuf

[Sebastiaan Couwenberg <sebastic@xs4all.nl>]

On 08/23/16 12:07, Sebastiaan Couwenberg wrote:
> On 08/23/16 11:32, Sebastiaan Couwenberg wrote:
> On s390x, alpha & sparc64 the build fails with:
>
>  ./.libs/libprotobuf.so: undefined reference to
> `google::protobuf::internal::NoBarrier_AtomicIncrement(long volatile*,
> long)'
>  ./.libs/libprotobuf.so: undefined reference to
> `google::protobuf::internal::NoBarrier_Store(long volatile*, long)'
>  ./.libs/libprotobuf.so: undefined reference to
> `google::protobuf::internal::NoBarrier_AtomicExchange(long volatile*,
> long)'
>  ./.libs/libprotobuf.so: undefined reference to
> `google::protobuf::internal::NoBarrier_Load(long const volatile*)'

I've just submitted a patch for the FTBFS on s390x in #835302, that 
should fix the last release architecture.

A similar patch is required for alpha & sparc64, which also lack 
explicit support for those architectures in atomicops.h. Fortunately 
these are not release architectures.

> For hurd-i386 a patch is needed to define PATH_MAX, and on kfreebsd-*
> the builds fails with:
>
>  google/protobuf/stubs/stringpiece_unittest.cc: In member function
> 'virtual void
> google::protobuf::{anonymous}::NonNegativeLenTest_NonNegativeLen_Test::TestBody()':
>
>  google/protobuf/stubs/stringpiece_unittest.cc:788:50: error:
> 'EXPECT_DEATH' was not declared in this scope
>     EXPECT_DEATH(StringPiece("xyz", -1), "len >= 0");
>                                                    ^
> On powerpcspe dh_strip fails with:
>
>  Not enough room for program headers

Ideally protobuf 3.0.0 gets fixed on the unofficial ports too, because 
2.6.x built successfully there before. But they are not blockers for 
this transition as s390x is.

Kind Regards,

Bas

--
 GPG Key ID: 4096R/6750F10AE88D4AF1
Fingerprint: 8182 DE41 7056 408D 6146  50D1 6750 F10A E88D 4AF1