Bug in complex::norm

[Gabriel Dos Reis]

nbecker@fred.net writes:

| The fact that it is not required to work doesn't mean we should
| intentionally choose an implementation that doesn't for no good
| reason.

Certainly. But your assertion that the implementation was wrong is,
err, simply wrong.

| complex<int> is very important for my work in Digital Signal
| Processing.

I'm not disputing that.  You said the implementation was wrong, but
there was no requirement that complex<> should work -- and I think we
should allow it to work.

-- Gaby