ARMv8 Ubuntu 18.04(64bit) Make error: requested alignment 64 is larger than 16

[Martin-Luther]

Hardware: RockPro64 with Rockchip RK3399 Hexa-Core (dual ARM Cortex A72 and quad ARM Cortex A53) 64-Bit Processor
ARCH: aarch64
OS: Ubuntu 18.04 (64 bit)
GCC: 7.3.0-27ubuntu1~18.04

When I enter : make -j $(nproc)

[ 34%] Building CXX object CMakeFiles/folly_base.dir/folly/experimental/coro/SharedMutex.cpp.o
[ 34%] Building CXX object CMakeFiles/folly_base.dir/folly/experimental/crypto/Blake2xb.cpp.o
[ 35%] Building CXX object CMakeFiles/folly_base.dir/folly/experimental/crypto/LtHash.cpp.o
[ 35%] Building CXX object CMakeFiles/folly_base.dir/folly/experimental/crypto/detail/MathOperation_AVX2.cpp.o
[ 35%] Building CXX object CMakeFiles/folly_base.dir/folly/experimental/crypto/detail/MathOperation_SSE2.cpp.o
[ 36%] Building CXX object CMakeFiles/folly_base.dir/folly/experimental/crypto/detail/MathOperation_Simple.cpp.o
In file included from /usr/local/src/folly/folly/experimental/crypto/LtHash.h:276:0,
                 from /usr/local/src/folly/folly/experimental/crypto/LtHash.cpp:17:
/usr/local/src/folly/folly/experimental/crypto/LtHash-inl.h: In member function ‘folly::crypto::LtHash<B, N>& folly::crypto::LtHash<B, N>::addObject(folly::ByteRange, Args&& ...)’:
/usr/local/src/folly/folly/experimental/crypto/LtHash-inl.h:343:53: error: requested alignment 64 is larger than 16 [-Werror=attributes]
   std::array<unsigned char, getChecksumSizeBytes()> h;
                                                     ^
/usr/local/src/folly/folly/experimental/crypto/LtHash-inl.h: In member function ‘folly::crypto::LtHash<B, N>& folly::crypto::LtHash<B, N>::removeObject(folly::ByteRange, Args&& ...)’:
/usr/local/src/folly/folly/experimental/crypto/LtHash-inl.h:361:53: error: requested alignment 64 is larger than 16 [-Werror=attributes]
   std::array<unsigned char, getChecksumSizeBytes()> h;
                                                     ^
cc1plus: all warnings being treated as errors
CMakeFiles/folly_base.dir/build.make:1167: recipe for target 'CMakeFiles/folly_base.dir/folly/experimental/crypto/LtHash.cpp.o' failed
make[2]: *** [CMakeFiles/folly_base.dir/folly/experimental/crypto/LtHash.cpp.o] Error 1
make[2]: *** Waiting for unfinished jobs....
/usr/local/src/folly/folly/experimental/crypto/detail/MathOperation_Simple.cpp: In static member function ‘static void folly::crypto::detail::MathOperation<E>::add(uint64_t, size_t, folly::ByteRange, folly::ByteRange, folly::MutableByteRange) [with folly::crypto::detail::MathEngine E = (folly::crypto::detail::MathEngine)1; uint64_t = long unsigned int; size_t = long unsigned int; folly::ByteRange = folly::Range<const unsigned char*>; folly::MutableByteRange = folly::Range<unsigned char*>]’:
/usr/local/src/folly/folly/experimental/crypto/detail/MathOperation_Simple.cpp:53:67: error: requested alignment 64 is larger than 16 [-Werror=attributes]
   alignas(kCacheLineSize) std::array<uint64_t, kValsPerCacheLine> results;
                                                                   ^~~~~~~
/usr/local/src/folly/folly/experimental/crypto/detail/MathOperation_Simple.cpp: In static member function ‘static void folly::crypto::detail::MathOperation<E>::sub(uint64_t, size_t, folly::ByteRange, folly::ByteRange, folly::MutableByteRange) [with folly::crypto::detail::MathEngine E = (folly::crypto::detail::MathEngine)1; uint64_t = long unsigned int; size_t = long unsigned int; folly::ByteRange = folly::Range<const unsigned char*>; folly::MutableByteRange = folly::Range<unsigned char*>]’:
/usr/local/src/folly/folly/experimental/crypto/detail/MathOperation_Simple.cpp:118:67: error: requested alignment 64 is larger than 16 [-Werror=attributes]
   alignas(kCacheLineSize) std::array<uint64_t, kValsPerCacheLine> results;
                                                                   ^~~~~~~
/usr/local/src/folly/folly/experimental/crypto/detail/MathOperation_Simple.cpp: In static member function ‘static void folly::crypto::detail::MathOperation<E>::clearPaddingBits(uint64_t, folly::MutableByteRange) [with folly::crypto::detail::MathEngine E = (folly::crypto::detail::MathEngine)1; uint64_t = long unsigned int; folly::MutableByteRange = folly::Range<unsigned char*>]’:
/usr/local/src/folly/folly/experimental/crypto/detail/MathOperation_Simple.cpp:182:67: error: requested alignment 64 is larger than 16 [-Werror=attributes]
   alignas(kCacheLineSize) std::array<uint64_t, kValsPerCacheLine> results;
                                                                   ^~~~~~~
cc1plus: all warnings being treated as errors
CMakeFiles/folly_base.dir/build.make:1206: recipe for target 'CMakeFiles/folly_base.dir/folly/experimental/crypto/detail/MathOperation_Simple.cpp.o' failed
make[2]: *** [CMakeFiles/folly_base.dir/folly/experimental/crypto/detail/MathOperation_Simple.cpp.o] Error 1
CMakeFiles/Makefile2:146: recipe for target 'CMakeFiles/folly_base.dir/all' failed
make[1]: *** [CMakeFiles/folly_base.dir/all] Error 2
Makefile:129: recipe for target 'all' failed
make: *** [all] Error 2

[Orvid]

Unfortunately, we don't have any systems we can really test this on to be able to fix it :(

If you want to compile this as-is, enabling C++17 mode (which supports over-aligned types) with a new enough compiler should work around this issue.

[JoeLoser]

Alternatively, if you rebase with master where we now don't treat every warning as an error, this should result in a warning but not be fatal.

[Martin-Luther]

Thanks Joe !
It seems to be better. I also tried to add -DCXX_STANDARD_REQUIRED=c++17 to enabled C++17 as Orvid suggested, but I don't think that it made any difference.

make -j $(nproc) EXTRA_CMAKE_OPTIONS='-DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc -DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++ -DCXX_STANDARD_REQUIRED=c++17'

At least, I am not stuck anymore, so thank you.

After some googling, I have read that this issue could be fixed if I install GCC 8.x or GCC 9. But I have not tested yet.

Thank you guys.