Special characters in mangled names

[Christophe de Dinechin]

> Two weeks ago we made some initial discussions about what special
> characters we could us as an "escape encoding" for mangled names
> (in cfront and cfront-like models, a double underscore is used,
> but it seems desirable to instead have a convention that cannot
> collide with C identifiers at all).

Is it really desirable? It did not cause a problem so-far, and there  
are occasions where being able to define a C++ function in C or  
assembly is useful. I had a case recently where I had to do that to  
fix a system library unsat (but since this is because of a competitor  
compiler's problem, I won't speak :-)


Christophe