Fix msgpack_cpp11 tests on 32-bit architectures

[jamessan]

I recently updated the Debian package for msgpack-c to 3.2.1 and found that there were a number of build failures. It appears that all of the 32-bit architectures (except m68k) failed with this error:

[20/122] /usr/bin/c++   -I../build-gtest/install/include -I../include -Iinclude -DMSGPACK_DEFAULT_API_VERSION=3 -std=c++11 -g -O2 -fdebug-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2    -Wall -Wextra -Wconversion -MD -MT test/CMakeFiles/msgpack_cpp11.dir/msgpack_cpp11.cpp.o -MF test/CMakeFiles/msgpack_cpp11.dir/msgpack_cpp11.cpp.o.d -o test/CMakeFiles/msgpack_cpp11.dir/msgpack_cpp11.cpp.o -c ../test/msgpack_cpp11.cpp
FAILED: test/CMakeFiles/msgpack_cpp11.dir/msgpack_cpp11.cpp.o 
/usr/bin/c++   -I../build-gtest/install/include -I../include -Iinclude -DMSGPACK_DEFAULT_API_VERSION=3 -std=c++11 -g -O2 -fdebug-prefix-map=/<<PKGBUILDDIR>>=. -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2    -Wall -Wextra -Wconversion -MD -MT test/CMakeFiles/msgpack_cpp11.dir/msgpack_cpp11.cpp.o -MF test/CMakeFiles/msgpack_cpp11.dir/msgpack_cpp11.cpp.o.d -o test/CMakeFiles/msgpack_cpp11.dir/msgpack_cpp11.cpp.o -c ../test/msgpack_cpp11.cpp
../test/msgpack_cpp11.cpp: In member function ‘virtual void MSGPACK_TIMESPEC_timespec_pack_convert_32bit_sec_Test::TestBody()’:
../test/msgpack_cpp11.cpp:1081:36: error: narrowing conversion of ‘4294967295’ from ‘long unsigned int’ to ‘__time_t’ {aka ‘long int’} [-Wnarrowing]
 1081 |     timespec val1{ 0xffffffffUL, 0 };
      |                                    ^
../test/msgpack_cpp11.cpp: In member function ‘virtual void MSGPACK_TIMESPEC_timespec_object_with_zone_32bit_sec_Test::TestBody()’:
../test/msgpack_cpp11.cpp:1097:36: error: narrowing conversion of ‘4294967295’ from ‘long unsigned int’ to ‘__time_t’ {aka ‘long int’} [-Wnarrowing]
 1097 |     timespec val1{ 0xffffffffUL, 0 };
      |                                    ^

Since this test is only run as part of the c++11 suite, I created this PR to ensure there's c++11 coverage (both 64 and 32-bit) for the CI runs.

Please let me know if you'd rather add two new jobs to the matrix rather than re-purposing existing ones.

[jamessan]

CI / Linux (1) shows the same failure now:

2020-01-03T02:11:52.2175428Z /home/runner/work/msgpack-c/msgpack-c/test/msgpack_cpp11.cpp:1085:20: error: constant expression evaluates to 4294967295 which cannot be narrowed to type '__time_t' (aka 'long') [-Wc++11-narrowing]
2020-01-03T02:11:52.2176133Z     timespec val1{ 0xffffffffUL, 0 };
2020-01-03T02:11:52.2176418Z                    ^~~~~~~~~~~~
2020-01-03T02:11:52.2177087Z /home/runner/work/msgpack-c/msgpack-c/test/msgpack_cpp11.cpp:1085:20: note: insert an explicit cast to silence this issue
2020-01-03T02:11:52.2177477Z     timespec val1{ 0xffffffffUL, 0 };
2020-01-03T02:11:52.2177736Z                    ^~~~~~~~~~~~
2020-01-03T02:11:52.2177974Z                    static_cast<__time_t>( )
2020-01-03T02:11:52.2178646Z /home/runner/work/msgpack-c/msgpack-c/test/msgpack_cpp11.cpp:1085:20: warning: implicit conversion changes signedness: 'unsigned long' to '__time_t' (aka 'long') [-Wsign-conversion]
2020-01-03T02:11:52.2178973Z     timespec val1{ 0xffffffffUL, 0 };
2020-01-03T02:11:52.2179210Z                  ~ ^~~~~~~~~~~~

[jamessan]

Feel free to push to my branch if you'd like to fix the issues in this PR. I added some commits to address the issues that came up.

[redboltz]

You added export CXX11="ON" to two of CI matrix.
But they are for C++03 test. So I don't want to add export CXX11="ON" there.

I expected that if I define CXX17="ON" in the CI matrix, then C++11 and C++17 tests are enabled. But it only enables C++17 tests.

I will apply your fix for test/msgpack_cpp11.cpp.
And will update from IF (MSGPACK_CXX11) to IF (MSGPACK_CXX11 OR MSGPACK_CXX17) at

msgpack-c/test/CMakeLists.txt

Line 58 in 4a94f83

IF (MSGPACK_CXX11)

instead of applying your .github/workflows/gha.yml fix.

This covers 64bit CXX11 at the pattern 2 and 32bit CXX11 at the pattern 3.

It that OK?

[jamessan]

You added export CXX11="ON" to two of CI matrix.
But they are for C++03 test. So I don't want to add export CXX11="ON" there.

Yeah, I was trying to avoid expanding the matrix more. I can understand not wanting to take that change.

It that OK?

Sounds good. I'll remove the gha.yml change from this branch.

[redboltz]

@jamessan , thank you. merged.