[cfe-dev] Providing C99/C11 features from C++11 and above

Louis Dionne via cfe-dev cfe-dev at lists.llvm.org
Tue Feb 12 13:37:48 PST 2019


In C++11 and above, C99/C11 features should be available even when using C++. For example, FLT_HAS_SUBNORM should be provided by <float.h>:

    $ cat <<EOF | clang++ -xc++ -std=c++11 -
    #include <float.h>

    #ifndef FLT_HAS_SUBNORM
    #   error "FLT_HAS_SUBNORM is missing"
    #endif
    EOF

This should also work when including `<cfloat>`. However, this currently:
- fails in -std=c++11 or above
- succeeds in -std=gnu++11 or above
- succeeds in -std=c11 or above
- succeeds in -std=gnu11 or above

If I understand correctly, the problem is that we currently have the following in clang/lib/Headers/float.h:

    #include_next <float.h> // pick up whatever the system provides

    ...
    #undef FLT_HAS_SUBNORM // undefine what the system provides
    ...

    // redefine our own version of the macro
    #if __STDC_VERSION__ >= 201112L || !defined(__STRICT_ANSI__)
        ...
    #  define FLT_HAS_SUBNORM __FLT_HAS_DENORM__
        ...
    #endif

The #if block checks whether we're compiling in C11 or above, or whether we're compiling in a non-strict dialect (basically some any gnuXXX dialect). However, it doesn't check for strictly conforming C++, and as a result we don't get the definitions. I'd like to know what's the right strategy for providing these definitions in C++, and whether there's an established way of doing this already. Naively, changing the #if to

    #if __STDC_VERSION__ >= 201112L || !defined(__STRICT_ANSI__) || (defined(__cplusplus) && __cplusplus >= 201103L)
        ...
    #  define FLT_HAS_SUBNORM __FLT_HAS_DENORM__
        ...
    #endif

seems to do the job. But I'm not very familiar with how Clang's headers are imbricked into libc++ and how they interact with per-platform headers, so I'm asking here.

Thanks,
Louis

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20190212/9b082af6/attachment.html>


More information about the cfe-dev mailing list