> From: Martin Sebor от имени Martin Sebor > To: [email protected] > Subject: Re: svn commit: r650933 - /stdcxx/trunk/src/num_put.cpp > >> URL: http://svn.apache.org/viewvc?rev=650933&view=rev >> Log: >> 2008-04-23 Farid Zaripov <[EMAIL PROTECTED]> >> >> STDCXX-880 >> * src/num_put.cpp [_RWSTD_DBL_SIZE == _RWSTD_LDBL_SIZE]: Call double >> type overload >> from long double type overload when sizeof (double) == sizeof (long >> double). > > What about __rw_fmat_infinite()? Do we need overloads for float > and long double or is the conversion from those two types safe > (as in doesn't trigger traps or floating exceptions) and > adequately efficient?
Yes we definitely need overloads of __rw_fmat_infinite(), but in the next release (when we would implement overloads for __rw_isxxx() for corresponding types). Now the only double versions are implemented, but __rw_fmat_infinite() isn't called for float and long double types because of __rw_isfinite() always returns true for float and long double (except the case when sizeof(double) == sizeof (long double)) . Farid.
