[boost] Re: [patch] quaternion/octonion compiler errors

From: Marcelo E. Magallon (mmagallo_at_[hidden])
Date: 2003-05-27 07:57:22

On Tue, May 27, 2003 at 12:22:27PM +0100, Reece Dunn wrote:

 > > I'm baffled.  The compiler says:
 >
 > [...]
 >
 > I got the same sort of messages. It is even harder when GCC says:
 > parse error before ';'!!!!

 Heh.  I meant I was baffled by the problem at large, not by the
 compiler message.  3.x is so much better in this respect!  At least the
 messages are readable now :-) (scary nightmares about those 1000+ lines
 long messages because of a misplaced semicolon)

 Anyway, I got it.  Context diff against CVS attached.

 > I think I understand why G++ 3.3 is getting confused. The
 > constructors appear near the top of the class definition, for example
 > quaternion< float >.
 > [...]
 > Try placing the constructor definitions at the end of the quaternion<
 > float > and quaternion< double > declerations.

 Yes, that sent me in the right direction, even if I still don't
 understand what exactly is going on.  At that point in the code the
 compiler knows:

    1. There's a template quaternion<T>
    2. There's a method T quaternion<T>::R_component_1()
    3. There's a forward declaration for a quaternion<float> and a
       quaternion<double> specialization.
    4. There's a function template for quaternion_type_converter<T,U>
    5. Said function calls R_component_1()
    6. Said function is instantiated for T=float, U=double.
    7. At this point the compiler complains that:

        float quaternion<float>::R_component_1()

       is not defined.  The compiler is right I guess, there's no
       definition for that function, only a declaration (see 3.)

 The patch makes a declaration for the function in 4. and pushes its
 definition to the end of the file.  That makes the compiler happy.  I
 don't think this will affect other compilers.

-- 
Marcelo