build error: initializer list would use explicit constructor on Linux GCC5.4

[nmgwddj]

Build environment: System: Linux x86_64 Compiler: GCC5 C++ STD: 14

error message:

include/sqlite_orm/sqlite_orm.h:6835:45: error: converting to 'std::tuple<const char*, int>' from initializer list would use explicit constructor 'constexpr std::tuple<_T1, _T2>::tuple(_U1&&, _U2&&) [with _U1 = const char*; _U2 = int; <template-parameter-2-3> = void; _T1 = const char*; _T2 = int]'
             return {s.c_str(), int(s.size())};

[fnc12]

are you sure that GCC5 supports C++14?

[nmgwddj]

https://en.cppreference.com/w/cpp/compiler_support/14

[nmgwddj]

I found this answer, not sure if it's related: https://stackoverflow.com/questions/47534938/explicit-constructors-and-nested-initializer-lists

[fnc12]

can you test with a different GCC version?

[nmgwddj]

can you test with a different GCC version?

Compiled successfully with gcc7 under Linux. But we may have to adapt to gcc5. I don't know if you have any suggestions for this. I can try to modify it locally for testing.

Configuration:
[settings]
arch=x86_64
arch_build=x86_64
build_type=Debug
compiler=gcc
compiler.cppstd=14
compiler.libcxx=libstdc++11
compiler.version=7
os=Linux
os_build=Linux
[options]
[build_requires]
[env]

[fnc12]

I use clang, other community guys use VC++. So yes we don't have any suggestions

[trueqbit]

sqlite_orm 1.7.1 is using a C++17 library feature here - see compiler support matrix entry on Improving std::pair and std::tuple [N4387]. The current dev branch is using a std::pair instead of std::tuple, which doesn't exhibit this problem, see also a test project on godbolt.

[nmgwddj]

sqlite_orm 1.7.1 is using a C++17 library feature here - see compiler support matrix entry on Improving std::pair and std::tuple [N4387]. The current dev branch is using a std::pair instead of std::tuple, which doesn't exhibit this problem, see also a test project on godbolt.

Thanks for your reply. I will try develop branch.

[trueqbit]

Latest dev branch is now testing gcc as well, so you have good chances things will work for you... Can we close the ticket?

[fnc12]

let's wait for @nmgwddj response first

[nmgwddj]

This is full log under Linux GCC5.5 log.txt

[fnc12]

looks like you are still getting error. Is that correct?

[nmgwddj]

Yes, but this time with a different error.

[fnc12]

which branch do you use? dev?

[nmgwddj]

which branch do you use? dev?

Yes, in the dev branch.

[fnc12]

are you using single header or no?

[nmgwddj]

I just switch the code to the dev branch and execute the following two CMake commands:

cmake -Bbuild -DCMAKE_BUILD_TYPE=Debug
# build
cmake --build build --config Debug

[fnc12]

weird things. I did the same on my mac but with clang and it worked well. Can you check the same with clang?

[trueqbit]

The error log that you attached is very strange:

error: ‘connection_ref’ does not name a type: connection_ref con;

The corresponding header file has been correctly included long before. Could you try the unmerged headers from not_single_header_include/sqlite_orm and find out why gcc 5 does not work?

[trueqbit]

@nmgwddj Can we close this ticket?

[nmgwddj]

We have upgraded gcc to 7 and temporarily closed this issue.