Re: Puzzling behaviour
- From: "qubit" <lauraeaves@xxxxxxxxx>
- To: <programmingblind@xxxxxxxxxxxxx>
- Date: Tue, 10 Nov 2009 22:46:27 -0600
I have an idea -- I wonder if your curly braces aren't out of phase... but here
are some thoughts:
It would help if you supplied the class with the headers of the 2 procs so we
can see what you are dealing with -- i.e., what are the names of the functions
and the class?
You copied just one line with some calls and conversions and refer vagely to an
error message. Are you saying this line appears in some proc, and that the
other proc is called on this line?
Note the hint the error message gives: local function needs to be called or
have its address taken. Back when I was working on C++ compilation, we had 3
types of scopes to deal with: global or namespace scope, nested scope (class
member), and local (entity declared inside function block).
If you have 2 functions the members of a class, they should be able to refer to
each other in the definition, even if function a calls b before b is declared.
This is because the function bodies aren't processed until after the class.
Borland's compiler may have some idiosyncrasies, but I believe it does handle
inline member function definitions correctly. So the next thing to look at is
whether the class is a locally declared class. Then it is still legal for it to
have member functions, but they have to be defined and used within the function
as they go out of scope otherwise. So is the class in question declared as a
local class?
The next analogous question is, did you accidently move your global or member
function from global scope into the body of a function? This can happen if the
curly braces are out of phase.
So check your scope nesting.
One other thing: if you move a global function f into a class C, so C::f is a
member function, and if you try to pass the address of that function to another
function, you need to declare the parameter of the second function so that it
has the right type. For example, suppose f0 is the function with the parameter
in question:
int f() { return 1; }
class C { public:
int f0( int(*pf)() ) { return pf(); }
};
note the function pointer pf.
Now move f() into class C so it is now a member:
class C { public:
int f() { return 1; }
int f0( int(*pf)() ) { return pf(); }
};
The declaration of f0 is bad. You will get loude complaints, because the
function call pf() is not a call to a member function. Try declaring the
following:
class C { public:
int f() { return 1; }
int f0( int(C::*pf)() ) { return this->pf(); }
};
Note the definition of f0 has to be modified to use the right type.
I hope thise musings are of use. I didn't expect them to get so long.
Happy hacking.
--le
----- Original Message -----
From: Ian D. Nichols
To: programmingblind@xxxxxxxxxxxxx
Sent: Tuesday, November 10, 2009 10:12 PM
Subject: Puzzling behaviour
Hi Listers,
I have a puzzling situation that I cannot figure out. This is C++, using a
Borland command-line compiler, Windows XP SP3, and JAWS 9.
The following line compiles just fine, unless it is one of the methods in a
class. Dlg1Proc is also a method of the class.
int Result = DialogBoxParam (hInstance, "Dialog1", hWnd, (DLGPROC)Dlg1Proc,
0);
It does not wrap in the source file.
If the 2 methods are in a class, the error message says that the local method
must be called or its address taken.
It would be much more convenient for me if it would work in the class. Does
anyone have any idea why it won't, and what my options are?
TIA.
Ian
Ian D. Nichols,
Toronto, Canada
Other related posts:
