Thanks for the notes! I'll update the way lit.cfg handles the std parameter.
================
Comment at: CMakeLists.txt:197
@@ -196,3 +196,3 @@
endif()
- if (LIBCXX_ENABLE_CXX0X AND LIBCXX_HAS_STDCXX0X_FLAG)
- list(APPEND LIBCXX_CXX_REQUIRED_FLAGS -std=c++0x)
+ if (LIBCXX_ENABLE_CXX1Y AND LIBCXX_HAS_STDCXX1Y_FLAG)
+ set(LIBCXX_STD_VERSION c++1y)
----------------
Dan Albert wrote:
> I think the case of LIBCXX_ENABLE_CXX1Y && !LIBCXX_HAS_STDCXX1Y_FLAG should
> raise an error, not fall back to C++11.
If that were to be the case then we should turn LIBCXX_ENABLE_CXX1Y off by
default. There is no need for CMake to fail because it can't find a c++1y flag
since libc++ builds as c++11.
Selfishly I want the c++1y to be enabled by default because I'm doing a lot of
work on the LFTS that requires c++1y and I don't want to specify it to run the
tests.
================
Comment at: test/lit.cfg:311
@@ +310,3 @@
+ assert std is not None
+ lit_config.note('inferred std as: \'-std={}\''.format(std))
+else:
----------------
Dan Albert wrote:
> It wouldn't be inferred, since it was defined by the user. I also don't think
> we need to be asserting on the line above. If std isn't defined either from
> the command line or from lit.site.cfg, we should probably just not pass any
> -std flag to the compiler and let it do the default. Might add a
> lit_config.note() stating that this is happening in that case.
I'm not sure if we should fall back to default. The majority of the test-suite
does not build unless its tested with >=c++11. I think we should instead fall
back to c++11.
http://reviews.llvm.org/D4329
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
