Strange MACRO Problem

Why following macro does not work?

#define removebrace(x) x

void Foo(int a, int b, char *txt, int d, int e);

main()
{


...
Foo(1, 2, removebrace(("hello", 5, 6)) );

...


}

I expected following expansion by compiler.

Foo(1, 2, "hello", 5, 6);

But it didnt work. When I see disassembly, only 3 arguments are pushed
into the stack, pointer to string "hello", value 2 and value 1 before
calling function 'Foo', instead of all 5 arguments.

However, if I do not use macro 'removebrace', all five arguments get
pushed to the stack before calling function 'Foo', and it works fine.

Any help, why MACRO removebrace does not work. I use MSVC 6.

Interestingly, this macro works fine in following statement

p = q * removebrace(a+b)

which results in

p = (q*a) + b

Karim

My previous post appeared to be lost so posting it again. Excuse me if
you receive two copy

re: Strange MACRO Problem

On Mon, 18 Jan 2004, Karim Thapa wrote:[color=blue]
>
> Why following macro does not work?
>
> #define removebrace(x) x[/color]

In other words, 'removebrace' is almost -- but not quite -- a no-op.
In particular, assuming 'foo' is not a #defined identifier,

removebrace(foo) ==> foo

for all 'foo'.
[color=blue]
> void Foo(int a, int b, char *txt, int d, int e);
>
> main()
> {
> ..
> Foo(1, 2, removebrace(("hello", 5, 6)) );[/color]

Since removebrace(("hello", 5, 6)) expands to ("hello, 5, 6), this
whole line expands to

Foo(1, 2, ("hello", 5, 6) );

This doesn't match the prototype, so the compiler should give you
a diagnostic error message.
[color=blue]
> I expected following expansion by compiler.
>
> Foo(1, 2, "hello", 5, 6);[/color]

Why on earth would you expect *that*?
[color=blue]
> But it didnt work. When I see disassembly, only 3 arguments are pushed
> into the stack, pointer to string "hello", value 2 and value 1 before
> calling function 'Foo', instead of all 5 arguments.[/color]

That's because your compiler has adequate constant-expression
optimization. It knows that since "hello" and 5 have no side-effects,
they can be simply discarded.
[color=blue]
> However, if I do not use macro 'removebrace'[/color]
[and remove both pairs of the extra parentheses][color=blue]
> , all five arguments get
> pushed to the stack before calling function 'Foo', and it works fine.[/color]

Naturally.
[color=blue]
> Any help, why MACRO removebrace does not work. I use MSVC 6.[/color]

It does work. It replaces removebrace(x) by x.
[color=blue]
> Interestingly, this macro works fine in following statement
>
> p = q * removebrace(a+b)
>
> which results in
>
> p = (q*a) + b[/color]

No, it results in

p = q * a+b

which has the same effect; but you've inserted a pair of
parentheses where they don't belong.

There is no C macro 'foo' such that 'foo((bar))' replaces to 'bar'
in general for all text 'bar', if that's what you're looking for.
What are you *really* trying to do, and why?

-Arthur

re: Strange MACRO Problem

"Karim Thapa" <karimthapa@yahoo.com> wrote in message
news:cda8d031.0401182313.42284a28@posting.google.c om...[color=blue]
> Why following macro does not work?
>
> #define removebrace(x) x
>
> void Foo(int a, int b, char *txt, int d, int e);
>
> main()
> {
>
>
> ..
> Foo(1, 2, removebrace(("hello", 5, 6)) );[/color]

This is equivalent to:

Foo ( 1,2, ("hello", 5, 6 ) );

which in turn is equivalent to:

Foo (1, 2, 6 );

Since the signature of Foo says expect 5 arguments, this one
will generate an error. Know/read about comma operator.

[..][color=blue]
>
> Any help, why MACRO removebrace does not work. I use MSVC 6.
>[/color]

This hasn't got anyting to do with MSVC 6.

Your problem solves if you replace

Foo(1, 2, removebrace(("hello", 5, 6)) );

with

Foo(1, 2, removebrace("hello", 5, 6) );

[..]

--
Vijay Kumar R Zanvar
My Home Page - http://www.geocities.com/vijoeyz/

re: Strange MACRO Problem

In article <bug6hb$h61cl$1@ID-203837.news.uni-berlin.de>,
"Vijay Kumar R Zanvar" <vijoeyz@hotpop.com> wrote:
[color=blue]
> "Karim Thapa" <karimthapa@yahoo.com> wrote in message
> news:cda8d031.0401182313.42284a28@posting.google.c om...[color=green]
> > Why following macro does not work?
> >
> > #define removebrace(x) x
> >
> > void Foo(int a, int b, char *txt, int d, int e);
> >
> > main()
> > {
> >
> >
> > ..
> > Foo(1, 2, removebrace(("hello", 5, 6)) );[/color]
>
> This is equivalent to:
>
> Foo ( 1,2, ("hello", 5, 6 ) );
>
> which in turn is equivalent to:
>
> Foo (1, 2, 6 );
>
> Since the signature of Foo says expect 5 arguments, this one
> will generate an error. Know/read about comma operator.
>
> [..][color=green]
> >
> > Any help, why MACRO removebrace does not work. I use MSVC 6.
> >[/color]
>
> This hasn't got anyting to do with MSVC 6.
>
> Your problem solves if you replace
>
> Foo(1, 2, removebrace(("hello", 5, 6)) );
>
> with
>
> Foo(1, 2, removebrace("hello", 5, 6) );[/color]

Most likely not, because now you are passing three arguments to a macro
that only expects one.

re: Strange MACRO Problem

Karim Thapa wrote:[color=blue]
> Why following macro does not work?
>
> #define removebrace(x) x
>
> void Foo(int a, int b, char *txt, int d, int e);
>
> main()
> {
> Foo(1, 2, removebrace(("hello", 5, 6)) );
> }[/color]

This snippet, BTW, is self-contradictory. Using implict int for the
retutn type of main is OK in C89 but not C99; leaving off the return value
is OK in C99 (where there is an implicit "return 0;") but not in C89.
[color=blue]
>
> I expected following expansion by compiler.
>
> Foo(1, 2, "hello", 5, 6);
>
> But it didnt work. When I see disassembly, only 3 arguments are pushed
> into the stack, pointer to string "hello", value 2 and value 1 before
> calling function 'Foo', instead of all 5 arguments.[/color]

All is as it should be. `x' is ("hello",5,6), not "hello",5,6.
[color=blue]
> Any help, why MACRO removebrace does not work. I use MSVC 6.[/color]

It works as it should.
[color=blue]
> Interestingly, this macro works fine in following statement
>
> p = q * removebrace(a+b)
>
> which results in
>
> p = (q*a) + b[/color]

No braces were there to be removed by removebace in your example above.
The parallel case is:
#define removebrace(x) x
int main(void)
{
int a = 1, b = 2, c = 3;
c = c * removebrace((a + b));
return 0;
}

expanding to:

int main()
{
int a = 1, b = 2, c = 3;
c = c * (a + b);
return 0;
}




--
Martin Ambuhl

re: Strange MACRO Problem

[..]
[color=blue][color=green]
> >
> > This hasn't got anyting to do with MSVC 6.
> >
> > Your problem solves if you replace
> >
> > Foo(1, 2, removebrace(("hello", 5, 6)) );
> >
> > with
> >
> > Foo(1, 2, removebrace("hello", 5, 6) );[/color]
>
> Most likely not, because now you are passing three arguments to a macro
> that only expects one.[/color]

True. I am committed to do one mistake per posting!! :-))

Thanks.

--
vijay-z

re: Strange MACRO Problem

Sorry, I made mistake while posting, it is actually

Foo(1, 2, removebrace("hello", 5, 6) );


Can some one explain now why macro removebrace does not
work as expected ?

I expected following expansion by compiler.

Foo(1, 2, "hello", 5, 6);

But it didnt work. When I see disassembly, only 3 arguments are pushed
into the stack, pointer to string "hello", value 1 and 2 before
calling function 'Foo', instead of all 5 arguments.

However, if I do not use macro 'removebrace', all five arguments get
pushed to the stack before calling function 'Foo'

Any help, why MACRO removebrace does not work. I use MSVC 6.

Karim

"Arthur J. O'Dwyer" <ajo@nospam.andrew.cmu.edu> wrote in message news:<Pine.LNX.4.58-035.0401190248310.20462@unix42.andrew.cmu.edu>...[color=blue]
> On Mon, 18 Jan 2004, Karim Thapa wrote:[color=green]
> >
> > Why following macro does not work?
> >
> > #define removebrace(x) x[/color]
>
> In other words, 'removebrace' is almost -- but not quite -- a no-op.
> In particular, assuming 'foo' is not a #defined identifier,
>
> removebrace(foo) ==> foo
>
> for all 'foo'.
>[color=green]
> > void Foo(int a, int b, char *txt, int d, int e);
> >
> > main()
> > {
> > ..
> > Foo(1, 2, removebrace(("hello", 5, 6)) );[/color]
>
> Since removebrace(("hello", 5, 6)) expands to ("hello, 5, 6), this
> whole line expands to
>
> Foo(1, 2, ("hello", 5, 6) );
>
> This doesn't match the prototype, so the compiler should give you
> a diagnostic error message.
>[color=green]
> > I expected following expansion by compiler.
> >
> > Foo(1, 2, "hello", 5, 6);[/color]
>
> Why on earth would you expect *that*?
>[color=green]
> > But it didnt work. When I see disassembly, only 3 arguments are pushed
> > into the stack, pointer to string "hello", value 2 and value 1 before
> > calling function 'Foo', instead of all 5 arguments.[/color]
>
> That's because your compiler has adequate constant-expression
> optimization. It knows that since "hello" and 5 have no side-effects,
> they can be simply discarded.
>[color=green]
> > However, if I do not use macro 'removebrace'[/color]
> [and remove both pairs of the extra parentheses][color=green]
> > , all five arguments get
> > pushed to the stack before calling function 'Foo', and it works fine.[/color]
>
> Naturally.
>[color=green]
> > Any help, why MACRO removebrace does not work. I use MSVC 6.[/color]
>
> It does work. It replaces removebrace(x) by x.
>[color=green]
> > Interestingly, this macro works fine in following statement
> >
> > p = q * removebrace(a+b)
> >
> > which results in
> >
> > p = (q*a) + b[/color]
>
> No, it results in
>
> p = q * a+b
>
> which has the same effect; but you've inserted a pair of
> parentheses where they don't belong.
>
> There is no C macro 'foo' such that 'foo((bar))' replaces to 'bar'
> in general for all text 'bar', if that's what you're looking for.
> What are you *really* trying to do, and why?
>
> -Arthur[/color]

re: Strange MACRO Problem

"Arthur J. O'Dwyer" <ajo@nospam.andrew.cmu.edu> wrote in message news:<Pine.LNX.4.58-035.0401190248310.20462@unix42.andrew.cmu.edu>...[color=blue]
> On Mon, 18 Jan 2004, Karim Thapa wrote:[/color]
....[color=blue]
> Foo(1, 2, ("hello", 5, 6) );
>
> This doesn't match the prototype, so the compiler should give you
> a diagnostic error message.
>[color=green]
> > I expected following expansion by compiler.
> >
> > Foo(1, 2, "hello", 5, 6);[/color]
>
> Why on earth would you expect *that*?
>[color=green]
> > But it didnt work. When I see disassembly, only 3 arguments are pushed
> > into the stack, pointer to string "hello", value 2 and value 1 before[/color][/color]
^^^^^^^^^^^^^^^^
it is strange,here.
the value of the expression ("hello",5,6) is 6,not "hello"??
i think value 6 would be pushed,not "hello".[color=blue][color=green]
> > calling function 'Foo', instead of all 5 arguments.[/color]
>
> That's because your compiler has adequate constant-expression
> optimization. It knows that since "hello" and 5 have no side-effects,
> they can be simply discarded.
>[color=green]
> > However, if I do not use macro 'removebrace'[/color]
> [and remove both pairs of the extra parentheses][/color]
....

re: Strange MACRO Problem

Karim Thapa wrote:[color=blue]
> Sorry, I made mistake while posting, it is actually
>
> Foo(1, 2, removebrace("hello", 5, 6) );
>
>
> Can some one explain now why macro removebrace does not
> work as expected ?[/color]

Because you gave it 3 arguments and it only takes 1:
[...][color=blue][color=green][color=darkred]
>>>#define removebrace(x) x[/color][/color][/color]

You should have gotten a compiler diagnostic. If you did not, turn the
warnings back on.


--
Martin Ambuhl

re: Strange MACRO Problem

Vijay Kumar R Zanvar wrote:

[color=blue][color=green]
>>#define removebrace(x) x[/color][/color]
[color=blue]
> Your problem solves if you replace
>
> Foo(1, 2, removebrace(("hello", 5, 6)) );
> with
> Foo(1, 2, removebrace("hello", 5, 6) );[/color]

Since when does supplying 3 arguments to a macro taking only one solve
anyone's problems?


--
Martin Ambuhl

re: Strange MACRO Problem

> Most likely not, because now you are passing three arguments to a macro[color=blue]
> that only expects one.[/color]

God! While doing something complex, I missed this silly point.

I have new problem now. I will post a new thread.

Thanks - Karim