TITLE: The maximal munch rule

PROBLEM: luj@helios.ecn.purdue.edu (John Lu)

I just moved into ++ world from C.  I don't see why AT&T CC 2.1 complains
for the following simple code(on a Sparc +1 running SunOS4.1)

friend void print(Foo&, ostream&=cout);
      	//????? must be   "ostream& = cout", i.e. must have space around "="
	//????? otherwise CC complains and quit


RESPONSE: Steve Clamage, TauMetric Corp (steve@taumet.com)

There are similar problems in C as well, for example:
	int i, j, k;
	int *p;
	j = i/*p;
	k = j /* intermediate result */ + 1;

OOPS!  This turns out to be equivalent to
	j = i + 1;

We intended to divide i by *p, but the lack of a space turned the '/*'
into the start of a comment, radically changing the meaning of the code.

The rule in C and C++ is that the longest sequence of characters which
could be a single token is taken to be that token, even if some other
interpretation would otherwise be possible.  (This is sometimes called
the "maximal munch" rule.)

I was bitten by this when I wanted the default value of a pointer
argument to be 0:
	foo(char*=0);
This produced bizarre error messages from the compiler!