Le 02/05/2018 à 14:55, Steven Watanabe via Boost a écrit :
AMDG
On 05/02/2018 05:53 AM, Raffi Enficiaud via Boost wrote:
Le 01/05/2018 à 23:23, Steven Watanabe via Boost a écrit :
On 05/01/2018 03:12 PM, Raffi Enficiaud via Boost wrote:
I am experiencing an issue with VS 2017 (15.6.6) while compiling boost.test on my agent. Traces below.
I do not recall having applied any change to this agent apart from installing C++/CLI support for CMake people.
<snip> build 01-May-2018 11:43:58 build 01-May-2018 11:43:58 call "..\..\..\bin.v2\standalone\msvc\msvc-14.1\msvc-setup.bat" >nul build 01-May-2018 11:43:58 cl /Zm800 -nologo @"..\..\..\bin.v2\libs\test\test\string_cast-test.test\msvc-14.1\dbg\thrd-mlt\utils-ts\string_cast-test.obj.rsp"
<snip>
Check msvc-setup.bat and make sure that it looks sane. If it's left over from a previous run, try deleting it and letting it be recreated.
Thanks Steven and Peter,
However, this happens on clean builds as well.
The only call that is done before the bootstrap.bat call is "call "C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\Tools\VsDevCmd.bat"
The msvc-setup.bat looks ok, as far as I understand it. However, the platform is set to x86 from the "b2 headers" part.
That's expected. If you use address-model=64, it will create another msvc-setup.bat.
a) Does it work if you tell Boost.Build to use the original scripts from visual studio directly? (using msvc : 14.1 : : <rewrite-setup-scripts>off ; ) b) Does the compiler work correctly if you run it manually?
I do b2 headers b2 toolset=msvc-14.1 (from libs/test/tests)
It used to work so far.
Thanks for all your feedback, Somehow adding address-model=64 at the header and build steps fixed everything. I do not know what I was compiling so far then :) Best, Raffi