[cfe-dev] merging of FunctionDecl's

[John McCall]

Douglas Gregor wrote:
> This is actually this issue. We should be mangling based on the  
> declaration we're given, not based on the canonical declaration (which  
> is the first declaration and, therefore, does not have the attribute).  
> Fixed here:
>   

Is there any compelling reason not to push this attribute onto the 
canonical declaration?  I'm not sure what exactly a canonical 
declaration is if it's not something you can always safely use without 
fearing loss of critical information.

John.