Attempt to fix VS build by defining noexcept

[dmah42]

No description provided.

[dmah42]

@BillyONeal any chance you can try this to see if it works for VS? I have no access to a VS machine.

[coveralls]

Coverage remained the same at 87.808% when pulling e515a66 on fixvscheck into 238e558 on master.

[coveralls]

Coverage increased (+0.5%) to 88.294% when pulling f2a1e44 on fixvscheck into 238e558 on master.

[BillyONeal]

I don't have super easy access to a VS2013 machine, and VS2015 supports conditional noexcept (so it works unmodified with the code in question), so it'll take a while for me to try. The build is already broken so if you don't want to wait for me to setup a VS2013 machine you may as well just merge it :)

[BillyONeal]

This should test _MSC_VER so that it uses conditional noexcept in 2015 or noexcept(false) cases will go to terminate() there.

[dmah42]

@BillyONeal ok. i know i fixed something, at least. but i think there's more to do with the noexcept macro and cmake.

[BillyONeal]

@dominichamon You may need separate macros; I know in the STL we have _NOEXCEPT for unconditional noexcept and _NOEXCEPT_OP(x) for conditional noexcept

[dmah42]

slowest. edit compile loop. ever.

[BillyONeal]

Yeah; it's a real shame that AppVeyor is so much slower here than travis-ci :(

[coveralls]

Coverage increased (+0.5%) to 88.294% when pulling 0948b38 on fixvscheck into 238e558 on master.

[coveralls]

Coverage increased (+0.5%) to 88.294% when pulling 0948b38 on fixvscheck into 238e558 on master.

[coveralls]

Coverage increased (+0.5%) to 88.294% when pulling f572d6b on fixvscheck into 238e558 on master.

[coveralls]

Coverage increased (+0.5%) to 88.294% when pulling 5b14d00 on fixvscheck into 238e558 on master.

[coveralls]

Coverage increased (+0.5%) to 88.294% when pulling f55649f on fixvscheck into 238e558 on master.

[dmah42]

i wish i could clean up the commit history here... @EricWF whenever you can, please persuade cla-bot that i'm not stealing your work :)

[coveralls]

Coverage increased (+0.5%) to 88.294% when pulling 1e2fca5 on fixvscheck into 238e558 on master.

[BillyONeal]

OK, I got a VS2013 machine to test with. How much do you guys really care about 2013? :)

FAILED: C:\PROGRA~2\MICROS~1.0\VC\bin\cl.exe   /nologo /TP -DHAVE_STD_REGEX -DHAVE_STEADY_CLOCK -D_CRT_SECURE_NO_WARNINGS -I..\include -I..\src /DWIN32 /D_WINDOWS  /GR /EHsc /W4 /D_DEBUG /MDd /Zi /Ob0 /Od /RTC1 /showIncludes /Fosrc\CMakeFiles\benchmark.dir\benchmark.cc.obj /Fdsrc\CMakeFiles\benchmark.dir\ /FS -c ..\src\benchmark.cc
C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\INCLUDE\xxatomic(229) : error C2664: 'std::_Atomic_address &std::_Atomic_address::operator =(const std::_Atomic_address &)' : cannot convert argument 1 from 'const char *' to 'void *'
        Conversion loses qualifiers
        C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\INCLUDE\xxatomic(227) : while compiling class template member function 'std::atomic<const char *>::atomic(_Ty *) throw()'
        with
        [
            _Ty=const char
        ]
        ..\src\benchmark.cc(118) : see reference to function template instantiation 'std::atomic<const char *>::atomic(_Ty *) throw()' being compiled
        with
        [
            _Ty=const char
        ]
        ..\src\benchmark.cc(118) : see reference to class template instantiation 'std::atomic<const char *>' being compiled
[14/37] Building CXX object src\CMakeFiles\benchmark.dir\colorprint.cc.obj
ninja: build stopped: subcommand failed.

[BillyONeal]

If you want I can make a separate PR with changes to fix VS2013

[dmah42]

yeah, i was seeing that on appveyor too. haven't managed to come up with a good solution yet that's clean.

if you have the time, go for it. otherwise i'll keep hacking on this.

[BillyONeal]

@dominichamon How about something like https://github.com/BillyONeal/benchmark/commit/9e0c515c93fada906c3bfe6eb190f20932a393d6 ?

[dmah42]

@BillyONeal right, that's something close to what i had. the const_cast bothered me, and the version check (instead of BENCHMARK_OS_WINDOWS or COMPILER_MSVC) but it may be that this is what we need to do.

you can land that independently of this PR, right?

[BillyONeal]

Yeah; I have to figure out how to fix the author on those commits first -- I had forgotten to setup my name correctly on the throwaway VS2013 VM where I made these changes :/

[EricWF]

Sorry I just saw your comments today. I'm OK with all of this.

[dmah42]

@EricWF no problem. can you convince cla-bot?

[EricWF]

@googlebot I'm Okay with these commits. How do I convince you?

[EricWF]

@googlebot I signed it.

[BillyONeal]

@dominichamon Do you want me to turn that commit into a PR or are you still pursuing an alternate solution?

[dmah42]

@BillyONeal please do. i could do it but you wrote the commit already :)