[Bug c++/11284] Bad code generation with "warning: the address ... will always be `true'"

[giovannibajo at libero dot it]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11284


giovannibajo at libero dot it changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mckelvey@maskull.com,
                   |                            |giovannibajo@libero.it


------- Additional Comments From giovannibajo at libero dot it  2003-06-22 16:53 -------
The warning is not a bug. The reduced snippet is:

---------------------------------
struct Bar
{
    Bar(bool );
};

struct Foo
{};

int main(void)
{
    Bar a(Foo());
    Bar b(a);

    return 0;
}
---------------------------------
pr11284.cpp: In function `int main()':
pr11284.cpp:12: warning: the address of `Bar a(Foo (*)())', will always be 
`true'


The problem is that the original submitter thinks that the first line in main() 
is creating an object of type Bar called "a". Instead, the line is declaring a 
function called "a", which returns an object of type Bar, and takes as argument 
a pointer to a function of type "Foo (*)()" (no parameters, returns a Foo). In 
other words, it's a function declaration.

So, when the second line is parsed (which is really an object definition), the 
compiler initializes Bar with function "a". While converting the function 
pointer to a boolean type, it emits a warning to notify the user that the 
result will always be true (a function pointer that comes from using a function 
name can never be NULL).

I keep the bug open because the testcase ICEs GCC 3.3. If someone wants to 
reduce it, we can check if it's a dupe.