Namespaces & main()

Namespaces & main()

[Gerald Gutierrez]

Hi all. I'm not very familiar with C++ namespaces yet so perhaps the
following may be correct behaviour. If so, please let me know.

If I try compiling the following :

namespace A { void main () { } }
namespace B { void main () { } }

egcs 1.1b gives me the following :

/tmp/ccNoTIxu.s: Assembler messages:
/tmp/ccNoTIxu.s:27: Fatal error: Symbol main already defined.

but if I renamed them to something else, like fmain(), then the code
compiles fine.

Alternatively, if I have the following :

namespace C {
    void main () { }
    namespace D {
        void main () { }
    }
}

I get :

/tmp/cc84q10E.s: Assembler messages:
/tmp/cc84q10E.s:27: Fatal error: Symbol main already defined.

as well. If I rename them to something else, the code compiles fine.

Is this correct C++ behaviour or is this something wrong with egcs ?

Re: Namespaces & main()

[Alexandre Oliva]

Gerald Gutierrez <gutier@intergate.bc.ca> writes:

> namespace A { void main () { } }
> namespace B { void main () { } }

> /tmp/ccNoTIxu.s: Assembler messages:
> /tmp/ccNoTIxu.s:27: Fatal error: Symbol main already defined.

> Is this correct C++ behaviour or is this something wrong with egcs ?

It's a bug in egcs.  main() is only reserved in the global namespace,
according to [basic.start.main]/3.  I've installed your testcase in
the testsuite.

-- 
Alexandre Oliva
mailto:oliva@dcc.unicamp.br mailto:oliva@gnu.org mailto:aoliva@acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil

Re: Namespaces & main()

[Branko Cibej]

Alexandre Oliva wrote:

> Gerald Gutierrez <gutier@intergate.bc.ca> writes:
>
> > namespace A { void main () { } }
> > namespace B { void main () { } }
>
> > /tmp/ccNoTIxu.s: Assembler messages:
> > /tmp/ccNoTIxu.s:27: Fatal error: Symbol main already defined.
>
> > Is this correct C++ behaviour or is this something wrong with egcs ?
>
> It's a bug in egcs.  main() is only reserved in the global namespace,
> according to [basic.start.main]/3.  I've installed your testcase in
> the testsuite.

Just FYI, I tried the following:

     namespace A { int main() {} }

and got

     namespace-main.cc: In function `int main()':
     namespace-main.cc:2: warning: control reaches end of non-void function `main()'

while the following:

     int main() {}

Did not produce a warning. All of this seems OK to me.

--
Branko Cibej   <branko.cibej@hermes.si>
HERMES SoftLab, Litijska 51, 1000 Ljubljana, Slovenia
phone: (++386 61) 186 53 49  fax: (++386 61) 186 52 70

Re: Namespaces & main()

[Todd Vierling]

On Wed, 14 Oct 1998, Branko Cibej wrote:

:      namespace A { int main() {} }

:      namespace-main.cc: In function `int main()':
:      namespace-main.cc:2: warning: control reaches end of non-void function `main()'

:      int main() {}
: 
: Did not produce a warning. All of this seems OK to me.

This part is correct.  The global "main" is explicitly defined in the C++
standard to return 0 if no return value is given.

-- 
-- Todd Vierling (Personal tv@pobox.com; Bus. todd_vierling@xn.xerox.com)

Re: Namespaces & main()

[Andreas Schwab]

Todd Vierling <tv@pobox.com> writes:

|> On Wed, 14 Oct 1998, Branko Cibej wrote:
|> 
|> :      namespace A { int main() {} }
|> 
|> :      namespace-main.cc: In function `int main()':
|> :      namespace-main.cc:2: warning: control reaches end of non-void function `main()'
|> 
|> :      int main() {}
|> : 
|> : Did not produce a warning. All of this seems OK to me.
|> 
|> This part is correct.  The global "main" is explicitly defined in the C++
|> standard to return 0 if no return value is given.

Unfortunately g++ doesn't implement that yet.  Thus you'll currently get a
random exit code in this case.

Andreas.

Re: Namespaces & main()

[Jason Merrill]

>>>>> Andreas Schwab <schwab@issan.informatik.uni-dortmund.de> writes:

 > |> This part is correct.  The global "main" is explicitly defined in the C++
 > |> standard to return 0 if no return value is given.

 > Unfortunately g++ doesn't implement that yet.

Huh?  Sure it does.

Jason

Re: Namespaces & main()

[Martin von Loewis]

> |> This part is correct.  The global "main" is explicitly defined in the C++
> |> standard to return 0 if no return value is given.
> 
> Unfortunately g++ doesn't implement that yet.  Thus you'll currently get a
> random exit code in this case.

??? On i486-pc-linux-gnu, when given

int main(){}

egcs-2.92.15 produces, with -O2 -S -fno-exceptions
-fomit-frame-pointer the output

main:
	xorl %eax,%eax
	ret

The bugs with 'main' in a namespace is a different issue. The name
'main' is not special inside a namespace: it is not the program entry
point, nor defaults its return value to 0.

Regards,
Martin

Re: Namespaces & main()

[Andreas Schwab]

Jason Merrill <jason@cygnus.com> writes:

|> >>>>> Andreas Schwab <schwab@issan.informatik.uni-dortmund.de> writes:
|> 
|>  > |> This part is correct.  The global "main" is explicitly defined in the C++
|>  > |> standard to return 0 if no return value is given.
|> 
|>  > Unfortunately g++ doesn't implement that yet.
|> 
|> Huh?  Sure it does.

The last time i checked (with the 981005 snapshot) it didn't.  Maybe that
changed in the mean time.

Andreas.

Re: Namespaces & main()

[Jason Merrill]

>>>>> Andreas Schwab <schwab@issan.informatik.uni-dortmund.de> writes:

 > Jason Merrill <jason@cygnus.com> writes:
 > |> >>>>> Andreas Schwab <schwab@issan.informatik.uni-dortmund.de> writes:
 > |> 
 > |>  > |> This part is correct.  The global "main" is explicitly defined 
 > |>  > |> in the C++ standard to return 0 if no return value is given.
 > |> 
 > |>  > Unfortunately g++ doesn't implement that yet.
 > |> 
 > |> Huh?  Sure it does.

 > The last time i checked (with the 981005 snapshot) it didn't.  Maybe that
 > changed in the mean time.

Just to be clear, I'm talking about falling off the end of ::main being
equivalent to a 'return 0;'.  g++ has supported this since before I started
working on it in 1993.

Jason

Re: Namespaces & main()

[Andreas Schwab]

Jason Merrill <jason@cygnus.com> writes:

|> Just to be clear, I'm talking about falling off the end of ::main being
|> equivalent to a 'return 0;'.  g++ has supported this since before I started
|> working on it in 1993.

Hmm, i tried it again, and it works as expected.  It seems like i forgot
to switch on my brain when i checked the last time. :-(

Andreas.