Updated mangling specs

[Daveed Vandevoorde]

mendell at ca.ibm.com wrote:
> 
> More questions:
> 
> This was one of the questions I wanted to ask Daveed.
> 
> The other one (maybe it's two) is:
> 
> Does this rule only distinguish between instantiation that would be
> ambiguous if both template definition were visible at the instantiation
> point? 

I believe that that is true.

> If so, do you know why the committee made this decision?

It grew out of discussions of what it meant for two template declarations
to refer to the same entity.  A number of people felt it was important
not to outlaw examples like the ones we have been discussing.  Not enough
people cared to defend the opposite situation.  Back then I was naive and
unexperienced ;-)  Oh, and I had no vote---don't shoot the messenger ;-)

> translation unit 1
> void f(char);
> void g() { f(1); }
> 
> translation unit 2
> void f(int);
> void h() { f(1); }
> 
> This is legal code, but I think is fairly obvious a blemish on the
> language. 

Hmmm, I cannot make up my mind about that.  Even when I transpose that
to a real module system, I see such situations as a possibility.

> External functions are usually declared in header files, so in
> general the programmer isn't going to know which f is being called and the
> inclusion of another header is going to change the behaviour of the
> program.  Paragraph 5 of 14.5.5.1 seems to me to be there solely to make
> this situation worse.  What am I missing?

I also find it to have little technical benefit.

It does guarantee us a job for life though <grin>.

	Daveed