https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88895
--- Comment #7 from Scott <scott.a.mayo at gmail dot com> --- While that's brilliant, 1) I don't lack for ways to code around the problem, but whether or not I do, g++ nonetheless has a problem here. 2) At a guess, the recursive solution is avoiding the warnings because the compiler is unable to figure out if the shifts are a problem, which probably means it can't optimize your solution. I could probably have hidden the size of the shifts with less code than your specializations uses: volatile unsigned int q = 56; //now the compiler can't assume the shift is a problem (T)v << q; or (T)v * (((uintmax_t)1 << 56)) but there goes hope of optimization, too. And for what I'm doing optimization is essential. I'm not looking for ways to code work around a warning. I'm just asking for a bugfix in the compiler. On 1/17/19 7:38 PM, msebor at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88895 > > --- Comment #6 from Martin Sebor <msebor at gcc dot gnu.org> --- > The C++ solution to this kind of problem is to use template specialization. > It's a been a while since I wrote any real C++ code but this seems to work and > avoids the warnings > > class PackRule > { > template <class T, unsigned N> > struct Packer; > public: > const unsigned short offset_; > const unsigned char width_; > > inline PackRule(const unsigned short offset, const unsigned char width) : > offset_(offset), width_(width) {} > > template <typename T> > inline void pack (void* to, const T v) const > { > unsigned char* tobyte = (unsigned char*)to + offset_; > Packer<T, sizeof (T)>::pack (tobyte, v); > } > > template<typename T> > inline T unpack (const void* from, T* output) const > { > const unsigned char* frombyte = (const unsigned char*)from + offset_; > return *output = Packer<T, sizeof (T)>::unpack (frombyte); > } > }; > > template <class T, unsigned N> > struct PackRule::Packer > { > static void pack (unsigned char *p, T x) > { > p[sizeof (T) - N] = (unsigned char)(x >> (N - 1) * __CHAR_BIT__); > Packer<T, N - 1>::pack (p, x); > } > > static T unpack (const unsigned char *p) > { > T x = (0xffU & (T)p[sizeof(T) - N]) << ((N - 1) * __CHAR_BIT__); > return x | Packer<T, N - 1>::unpack (p); > } > }; > > template <class T> > struct PackRule::Packer<T, 0> > { > static void pack (unsigned char*, T) { } > static T unpack (const unsigned char*) { return T (); } > }; > > int main () > { > PackRule p(0, 4); > unsigned char buf[4]; > p.pack (buf, (unsigned int)444); > unsigned int t; > p.unpack (buf, &t); > } >
