Some 32-bit platforms need explicit linking to libatomic for 8-byte atomics support

[barracuda156]

Some 32-bit platforms (ppc, arm, likely mips) need explicit linking to libatomic here:

[100%] Linking CXX executable FuzzFailTest
cd /opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_devel_CLI11/CLI11/work/build/tests && /opt/local/bin/cmake -E cmake_link_script CMakeFiles/FuzzFailTest.dir/link.txt --verbose=ON
/opt/local/bin/g++-mp-13 -pipe -Os -DNDEBUG -I/opt/local/include -D_GLIBCXX_USE_CXX11_ABI=0 -mmacosx-version-min=10.6 -Wl,-search_paths_first -Wl,-headerpad_max_install_names -L/opt/local/lib -Wl,-headerpad_max_install_names CMakeFiles/FuzzFailTest.dir/FuzzFailTest.cpp.o CMakeFiles/FuzzFailTest.dir/__/fuzz/fuzzApp.cpp.o -o FuzzFailTest  -Wl,-rpath,/opt/local/lib /opt/local/lib/libCatch2Main.a /opt/local/lib/libCatch2.a 
Undefined symbols:
  "___atomic_store_8", referenced from:
      __ZNSt17_Function_handlerIFbRKSt6vectorISsSaISsEEEZN3CLI3App8add_flagISt6atomicIdELNS6_6detail7enablerE0EEEPNS6_6OptionESsRT_SsEUlS4_E_E9_M_invokeERKSt9_Any_dataS4_ in fuzzApp.cpp.o
      __ZNSt17_Function_handlerIFbRKSt6vectorISsSaISsEEEZN3CLI3App10add_optionISt6atomicIxESA_LNS6_6detail7enablerE0EEEPNS6_6OptionESsRT_SsEUlS4_E_E9_M_invokeERKSt9_Any_dataS4_ in fuzzApp.cpp.o
      __ZNSt17_Function_handlerIFbRKSt6vectorISsSaISsEEEZN3CLI3App10add_optionISt6atomicIyESA_LNS6_6detail7enablerE0EEEPNS6_6OptionESsRT_SsEUlS4_E_E9_M_invokeERKSt9_Any_dataS4_ in fuzzApp.cpp.o
  "___atomic_load_8", referenced from:
      __ZNSt17_Function_handlerIFSsvEZN3CLI3App10add_optionISt6atomicIxES5_LNS1_6detail7enablerE0EEEPNS1_6OptionESsRT_SsEUlvE_E9_M_invokeERKSt9_Any_data in fuzzApp.cpp.o
      __ZNSt17_Function_handlerIFSsvEZN3CLI3App10add_optionISt6atomicIyES5_LNS1_6detail7enablerE0EEEPNS1_6OptionESsRT_SsEUlvE_E9_M_invokeERKSt9_Any_data in fuzzApp.cpp.o
ld: symbol(s) not found
collect2: error: ld returned 1 exit status
make[2]: *** [tests/FuzzFailTest] Error 1

[henryiii]

Is this only happening on the new fuzzing test?

[barracuda156]

Is this only happening on the new fuzzing test?

Looks like yes, at least on my system.

[henryiii]

OK, that should be pretty easy to fix.

[barracuda156]

@henryiii That would be great.

[barracuda156]

Passing -latomic locally via configure.cxxflags-append -Wl,-latomic (for some reason a more standard configure.ldflags-append -latomic failed in this case) fixed the build.
All tests pass:

--->  Testing CLI11
Executing:  cd "/opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_devel_CLI11/CLI11/work/build" && /usr/bin/make test 
Running tests...
/opt/local/bin/ctest --force-new-ctest-process 
Test project /opt/local/var/macports/build/_opt_PPCSnowLeopardPorts_devel_CLI11/CLI11/work/build

CLI11 information:
  C++ standard: 11
  __has_include: yes
  No optional library found

      Start  1: HelpersTest
 1/21 Test  #1: HelpersTest ......................   Passed    0.03 sec
      Start  2: ConfigFileTest
 2/21 Test  #2: ConfigFileTest ...................   Passed    0.10 sec
      Start  3: OptionTypeTest
 3/21 Test  #3: OptionTypeTest ...................   Passed    0.05 sec
      Start  4: SimpleTest
 4/21 Test  #4: SimpleTest .......................   Passed    0.02 sec
      Start  5: AppTest
 5/21 Test  #5: AppTest ..........................   Passed    0.09 sec
      Start  6: SetTest
 6/21 Test  #6: SetTest ..........................   Passed    0.03 sec
      Start  7: TransformTest
 7/21 Test  #7: TransformTest ....................   Passed    0.03 sec
      Start  8: CreationTest
 8/21 Test  #8: CreationTest .....................   Passed    0.03 sec
      Start  9: SubcommandTest
 9/21 Test  #9: SubcommandTest ...................   Passed    0.05 sec
      Start 10: HelpTest
10/21 Test #10: HelpTest .........................   Passed    0.04 sec
      Start 11: FormatterTest
11/21 Test #11: FormatterTest ....................   Passed    0.02 sec
      Start 12: NewParseTest
12/21 Test #12: NewParseTest .....................   Passed    0.02 sec
      Start 13: OptionalTest
13/21 Test #13: OptionalTest .....................   Passed    0.01 sec
      Start 14: DeprecatedTest
14/21 Test #14: DeprecatedTest ...................   Passed    0.01 sec
      Start 15: StringParseTest
15/21 Test #15: StringParseTest ..................   Passed    0.02 sec
      Start 16: ComplexTypeTest
16/21 Test #16: ComplexTypeTest ..................   Passed    0.04 sec
      Start 17: TrueFalseTest
17/21 Test #17: TrueFalseTest ....................   Passed    0.02 sec
      Start 18: OptionGroupTest
18/21 Test #18: OptionGroupTest ..................   Passed    0.04 sec
      Start 19: EncodingTest
19/21 Test #19: EncodingTest .....................   Passed    0.01 sec
      Start 20: TimerTest
20/21 Test #20: TimerTest ........................   Passed    0.24 sec
      Start 21: link_test_2
21/21 Test #21: link_test_2 ......................   Passed    0.01 sec

100% tests passed, 0 tests failed out of 21

Total Test time (real) =   0.91 sec

[henryiii]

Could you see if #1000 works?

[barracuda156]

@henryiii Apparently yes, it builds with this patch without manual flag passing.

[barracuda156]

@henryiii Is it possible to make a minor release incorporating recent fixes? At least 30ead5d (but it will be nice to have libatomic one too).
I have been asked by the port maintainer about that.

[henryiii]

Yes, planned.