Steven Watanabe wrote:
Richard Smith <richard <at> ex-parrot.com> writes:
I've just found what I believe to be a new regression on the boost RC_1_34_0 branch relative to the 1.33.1 release. The example below is a simplified extract of some real world where the behaviour of boost::none silently changes.
Doesn't internal linkage in headers have the same problem as unnamed namespaces? Is there some reason we can't use
[...] This works for me: namespace boost { namespace detail { struct none_helper {}; } typedef int (*none_t)(detail::none_helper); inline int none(detail::none_helper) { return 0; } } (Which is your version with the obvious typos and warnings corrected.) That removes the regression I was describing. But I don't have ready access to one of the compilers (Borland or VC7.1 ?) that exhibits problems with precompiled headers. Richard Smith