Re: [ublas] Problem with "has_trivial_contructor" trait for std::complex types

Subject: Re: [ublas] Problem with "has_trivial_contructor" trait for std::complex types
From: Neal Becker (ndbecker2_at_[hidden])
Date: 2011-12-16 07:50:53

Maharshi, Atul wrote:

> Hi,
> 
> The "has_trivial_contructor" and "has_trivial_destructor" traits as currently
> defined as:
> 
>         template<typename FLT>
>         struct has_trivial_constructor<std::complex<FLT> > : public
>         boost::true_type {};
> 
>         template<typename FLT>
>         struct has_trivial_destructor<std::complex<FLT> > : public
>         boost::true_type {};
> 
> This can be problematic if the FLT type has non-trivial contructor/destructor
> (see, for example, SystemC fixed point types -
> http://www.lysium.de/docs/systemc-2.2/docs/html/classsc__dt_1_1sc__fixed.html)
> 
> Perhaps, the "right" way to do define the traits is as follows:
> 
>         template<typename FLT>
>         struct has_trivial_constructor<std::complex<FLT> > : public
>         boost::has_trivial_constructor<FLT> {};
> 
>         template<typename FLT>
>         struct has_trivial_destructor<std::complex<FLT> > : public
>         boost::has_trivial_destructor<FLT> {};
> 
> 
> I understand the C++ standard states that the "effect of instantiating the
> template complex for any type other than float, double or long double" is
> unspecified. But even for these types, one could assume that a library could
> use a non-trivial default constructor.
> 
> Please let me know your thoughts.
> 
> Thanks,
> -Atul

Untested, but I like the above suggestion.