eppstein@garfield.columbia.edu (David Eppstein) (12/15/87)
I want to use automatic conversion of a certain class to strings, but it
seems to screw up other stuff around it. This is in
#ident "@(#)cfront:CC 1.11"
and the error messages generated in the following program are:
CC test.c:
"test.c", line 23: error: two initializers for baz() argument x
"test.c", line 23: error: bad initializer type const_char_star
for x ( int expected)
2 errors
The program itself:
------------ ------------
typedef const char * const_char_star;
class string {
public:
const char * p;
operator const_char_star() { return p; }
};
int strlen(const char *);
int baz(int = -1);
int foo()
{
string bar;
bar.p = "bar";
return strlen(bar);
}
int baz(int x = -1)
{
return x;
}
--
David Eppstein eppstein@garfield.columbia.edu Columbia U. Computer Scienceshopiro@alice.UUCP (12/17/87)
In article <5192@columbia.edu>, eppstein@garfield.columbia.edu.UUCP writes:
< I want to use automatic conversion of a certain class to strings, but it
< seems to screw up other stuff around it. This is in
< #ident "@(#)cfront:CC 1.11"
< and the error messages generated in the following program are:
<
< CC test.c:
< "test.c", line 23: error: two initializers for baz() argument x
< "test.c", line 23: error: bad initializer type const_char_star
< for x ( int expected)
< 2 errors
<
< The program itself:
<
< ...
<
< int baz(int = -1);
< ...
< int baz(int x = -1)
< {
< return x;
< }
< --
< David Eppstein eppstein@garfield.columbia.edu Columbia U. Computer Science
Of the two error messages you got, one should be there. A default
initializer may only appear once, in the declaration of a function.
The reason is that if two initializers were given for the same argument,
(one in the declaration and one in the definition) and they had different
values, we would interpret that as an error. But it is hard for the
cfront to check whether two expressions have the same value, since
certain constant expressions (e.g., floating point) are evaluated by
the underlying C compiler. So we only allow one initializer.
The other error message should not be there, and it isn't in the latest
internal version of cfront. I don't know when that version will make it
out the door, but I hope it will be soon.
--
Jonathan Shopiro
AT&T Bell Laboratories, Murray Hill, NJ 07974
research!shopiro (201) 582-4179