is this a string bug?

[Joe Buck]

> In chapter 5 of the new Stroustrup edition there is the program
> shown below.  It produces:

....

It appears that the problem is with the line

>     Pair p = {s, 0};

I think the code to handle this was originally put in as a GNU extension
to C; also I'm fairly sure that this form is a recent addition to C++.
Right?  (Because of this I had always thought of this form as a GNU
extension and so always avoided it).

Anyway, it seems to be malfunctioning in that it appears that a shallow
copy is being done to initialize the string field of p rather than using
the copy constructor.

If you write a constructor for Pair, and replace that line with

	Pair p(s,0);

the program will work correctly (Pair has to also have a default
constructor).  So it is not a bug in the string class, but in the
handling of aggregate initialization.