Re: [GD-Windows] VS 2003 generating duplicate COMDAT for template instantiation?

[Kent Q. <ken...@co...>]

Well, my template-fu is a bit underused on my current project, so I 
can't look at this and give you a solution.

But try this: chop this code down into the smallest compileable file you 
can, and then feed it to GCC 4 and maybe the Comeau computing compiler 
test page:
    http://www.comeaucomputing.com/tryitout/

I've found in the past that when I've had tricky template troubles, one 
or the other of those compilers will generate some message that will 
give me a clue. I know this is failing on the link, but if you turn on 
maximum warnings, I wouldn't be surprised to have one of these compilers 
spit out a warning or an error.

    Kent



Jon Watte wrote:
> I'm trying to generate a SAS parser/binder using just a little bit of 
> template programing, and I'm running into the following link error (I'm 
> using VS 2003). I would like any help I can get in how to work around 
> this problem. I got it when I was reinterpret_cast-ing from type 
> Foo::*Bar[N] to Foo::*Bar[], but I worked around that by explicitly 
> passing in the array size. However, here there are no skanky casts 
> involved; this should "just work." (And I've tried with and without the 
> inline Access() function).
>
>
> Win32DX9Graphics.obj : fatal error LNK1179: invalid or corrupt file: 
> duplicate COMDAT 
> '?Access@?$FieldAccessor@USasAmbientLight@impl_DX9@@VFastTriple@pit@@$GA@A@A@@impl_DX9@@QAEABVFastTriple@pit@@ABUSasAmbientLight@2@@Z'
>
> Looking at the templates FieldAccessor, here is the code:
>
>
> template<typename Container, typename Type, Type Container::*Field>
> class FieldAccessor {
>   public:
>     typedef Container ContainerType;
>     typedef Type FieldType;
>     inline Type const &Access(Container const &c) { return c.*Field; }
> };
>
>
> Here are the two cases where FieldAccessor<> is referenced for the 
> SasAmbientLight:
>
>
>       //  dispatch the child
>       if (!wcscmp(path, L"color")) {
>         setter_ = PIT_SYSTEMNEW0("ValueSetter") ValueSetter<
>             FieldAccessor<SasAmbientLight, pit::FloatTriple, 
> &SasAmbientLight::color> >
>             (this, core_->ambientlight_[l]);
>         return true;
>       }
>       return false;
>
> ...
>
>       //  dispatch the child, then return an iterator on that
>       if (!wcscmp(path, L"color")) {
>         setter_ = PIT_SYSTEMNEW0("OuterArraySetter") OuterArraySetter<
>             FieldAccessor<SasAmbientLight, pit::FloatTriple, 
> &SasAmbientLight::color> >
>             (this, core_->ambientlight_, &core_->numambientlights_.get());
>         return true;
>       }
>       return false;
>
>
>
> pit::FloatTriple is what you'd expect, as is struct SasAmbientLight:
>
> struct SasAmbientLight {
>   pit::FloatTriple      color;
> };
>
>
>
> Btw: upgrading to 2005 isn't the preferred solution right now.
>
>
>   


-- 
------------------------------------------------------------
Kent Quirk           I'm making a game about global warming.
Game Architect                        Track the progress at:
CogniToy                http://www.cognitoy.com/meltingpoint