Re: wmemchr on MSVC

[Bruno Haible]

Eric Blake wrote:
> letting MSVC provide only an inline function will break the 
> ability to pass around a function pointer for wmemchr.

No, it will only break the ability to compare such function pointers
with ==.

> I'm surprised  that letting the inline function get by still passes the
> signature checks in the unit tests.

Yes, I was surprised too, and it even passes the signature check in C++ mode.
But it does pass the tests, therefore I don't see the need to enable the
replacement.

It's similar to stat() and lstat() on some Unix platforms, which sometimes
are defined as inline functions in terms of a function xstat() or so.
If it's POSIX compliant that glibc defines stat() and lstat() as inline
functions, it must also be POSIX compliant to define wmemchr() as an inline
function.

Bruno