I don't know if this is relevant or adds to your conversation but it seems a few things I see are related:
* I fixed the always inline warnings for libcxx so they don't appear anymore for cl based compiles at least. I haven't seen any warnings at least related to libcxx using g++/clang++/cl.exe for that particular error now once that patch is applied. * When compiling with cl.exe in the win32 configuration I get these errors: c:\libcxx\include\support/win32/support.h(98): error C3861: '_BitScanReverse64': identifier not found c:\libcxx\include\support/win32/support.h(113): error C3861: '_BitScanForward64': identifier not found Because neither function seems to be available for Win32 configurations, but if you compile using for Win64 configuration it compiles fine. I need to add some non intrinsic support for cl for the 32 bit case I am thinking for cl and the MSVCRT. * I was going to propose a patch to add "#pragma clang system_header" to a "\libcxx\include\support\win32\support.h" and other such files in the support folders. Is there any reason why I shouldn't. PS Of the wall, probably not thought out question: shouldn't the system header indicator be hereditary. i.e. any file that is a system header that includes another file, then that should be seen as a system header too. I'm not asking anyone to make that change, but is it conceptually sound? Thanks
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
