Re: [Gnash-dev] ACT module; white box testing and aspect templates

[Eric Hughes]

Here's the operational reply:

At 08:17 AM 7/18/2007, strk wrote:
> cygnal/ACT/Handle.hpp:122: error: invalid default argument for a template 
> template parameter

Null_Aspect_1 should have a namespace qualifier attached to it.  Use 
aspect::Null_Aspect_1.

The fact that this compiles under MSVC is most likely evidence of a 
compiler defect, either in MSVC or gcc, or else a standards defect.  As I 
recall the code does not instantiate that particular class with its default 
template; it is instantiated with another class's default aspect (in this 
case).  As a rule, templates delay their instantiation, so it seems 
arguable that an unknown default argument ought to be legal.  On the other 
hand, a default argument might be argued to be an independent name, which 
must be legal at the point of definition; this is what gcc seems to be 
claiming.  MSVC, on the other hand, may be read to claim that a default 
parameter value only comes into scope at the point of 
instantiation.  Stroustrup C.13.8 is the relevant section; I can't find 
anything definitive in it.  I'm not going to bother reading the C++ 
standard on this.

While interesting, all of the above matters little, because the default 
template parameter ought to work in other contexts.

Eric