building fails with '-DENABLE_TESTING=ON'

[xtcyclist]

Please check the FAQ documentation before raising an issue

Describe the bug (required)

(1) BENCHMARK_RELATIVE(cord_1k_mix, iters) and BENCHMARK_RELATIVE(icord_1k_mix, iters) block the build with the following errors:

/usr/include/x86_64-linux-gnu/bits/string_fortified.h:34:33: error: ‘void* __builtin_memcpy(void*, const void*, long unsigned int)’ reading between 1 and 1024 bytes from a region of size 0 [-Werror=stringop-overread]

(2) src/common/datatypes/test/ValueTest.cpp

/usr/bin/ld.bfd: CMakeFiles/value_test.dir/ValueTest.cpp.o: in function `nebula::Value_DecodeEncode_Test::TestBody()':
ValueTest.cpp:(.text+0x318d4): undefined reference to `unsigned int apache::thrift::Cpp2Ops<nebula::Value, void>::write<apache::thrift::CompactProtocolWriter>(apache::thrift::CompactProtocolWriter*, nebula::Value const*)'

Your Environments (required)

• OS: Linux nebula-dev-84 5.4.0-105-generic
• Compiler: 11.1.0
• CPU: Intel(R) Xeon(R) Platinum 8352Y CPU @ 2.20GHz
• Commit id: 7ce81ee

How To Reproduce(required)

Steps to reproduce the behavior:

1. Step 1

rm -rf build
mkdir build
cd build
cmake
-DENABLE_TESTING=ON
-DCMAKE_BUILD_TYPE=Release
..

make -j32
make install

3. Step 2
4. Step 3

Expected behavior

Additional context

[xtcyclist]

open happens with the release, not debug mode. seems to be some issue with the compiler.

[xtcyclist]

For the first issue with __builtin_memcpy, passing a character sequence like "abcdef", as in the benchmark, into cord's or icord's writer() triggers the compiler to warn possible problems like stringop-overread or array out-of-bound.

Can be fixed by first converting the character sequence into a std::string, and then pass the std::string to the writer().

BTW, the code looks fine to me. it's just that gcc-11 warns of potential problems.

[xtcyclist]

For the second problem, the definitions for the class apache::thrift::Cpp2Ops's methods are written in some header files that are not included in the tests.