Re: operators.hpp and MSVC bugs - now I've seen everything

From: boost_at_[hidden]
Date: 2001-02-25 19:17:17

Never mind. After finding 3 MSVC template bugs today, I downloaded 
the Intel 5.0 compiler, and I must say that so far I like what I see 
very much!

It compiles quickly and correctly. It also confirms that the issue I 
mentioned below is a MSVC bug - I don't need a confirmation.

If anyone in the Visual C++ team is reading this list - I say watch 
out - you've got serious competition! My goal right now is to buy the 
Intel compiler and hopefully never have to deal with MSVC again!

Paul Hollingsworth
http://PaulHollingsworth.com

--- In boost_at_y..., boost_at_P... wrote:
> Can someone just confirm that this is a compiler bug (I'm running 
> MSVC sp4 I think), for my own sanity?
> 
> If a local class called "B" is in the same scope, then B ends up 
> being used instead of the template parameter "B" as defined in 
> operators.hpp.
> 
> I would have thought that the template name should override the non-
> local class definition though.
> 
> I suppose I'll just clock it down to a few more hours wasted 
> discovering another Microsoft compiler bug:
> 
> #include <boost/operators.hpp>
> 
> // Uncomment the following class definition to get
> // "error C2676: binary '*' : 'class point' does not define this
> // operator or a conversion to a type acceptable to the predefined 
> operator"
> 
> // struct B {};
> 
> class point
> : boost::addable<point,
> boost::multipliable2<point, float> >
> {
> public:
> 	point operator+=(const point &rhs)
> 	{
> 		return *this;
> 	}
> 	point operator*=(float rhs)
> 	{
> 		return *this;
> 	}
> };
> 
> int main()
> {
> 	point a;
> 	point b;
> 	a * 5.0f;
> 	return 0;
> }