G++/GCC not detetcing stupid errors in code.

G++/GCC not detetcing stupid errors in code.

[Laurent Dufréchou]

Hi there,

I’m using g++ for the first time via gcc 4.1.2.
I’ve written a little code (first C++ oriented and because guys with which I
work come from C, it is slightly modified to compile with G++ and gcc)

>>>>>>>>>>>>>>>>>>> Code >>>>>>>>>>>>>>>>>>>>>>>>>>>>
 
//#include <iostream>
#include <stdio.h>

//using namespace std;

int main()

{

                int toto;

                short tata;

                toto = 1111118;

                tata = toto;  //error 1!

                //cout << toto << "tata:" << tata << endl;

                printf("%d %i",toto,tata);

                //return 0; //error 2 because no return

}

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

I i compile with gcc (cc), it detects no int returned error (OK good point)
But the downcast (tata = toto) doesn’t raise any warning!

If I use G++ the downcast is not detected and even more the return is not
also detected (!)

I was thining that Gcc was really strict and I choose it because I was
thinking that, I’m quite disappointed :/
I’ve tried –Wall –Wextra –pedantic + a ton of other switch with no luck…

Do i miss something ? 

How can I put Gcc in “user is stupid I need to check all” mode?


Laurent

Re: G++/GCC not detetcing stupid errors in code.

[Sven Eschenberg]

You are looking for -Wconversion I guess.

Implicit conversion are a nice thing to have ;-).

Regards

-Sven


Laurent Dufréchou schrieb:
> Hi there,
>
> IÂ’m using g++ for the first time via gcc 4.1.2.
> IÂ’ve written a little code (first C++ oriented and because guys with which I
> work come from C, it is slightly modified to compile with G++ and gcc)
>
>   
>>>>>>>>>>>>>>>>>>>> Code >>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>                                         
>  
> //#include <iostream>
> #include <stdio.h>
>
> //using namespace std;
>
> int main()
>
> {
>
>                 int toto;
>
>                 short tata;
>
>                 toto = 1111118;
>
>                 tata = toto;  //error 1!
>
>                 //cout << toto << "tata:" << tata << endl;
>
>                 printf("%d %i",toto,tata);
>
>                 //return 0; //error 2 because no return
>
> }
>
>   
>
> I i compile with gcc (cc), it detects no int returned error (OK good point)
> But the downcast (tata = toto) doesnÂ’t raise any warning!
>
> If I use G++ the downcast is not detected and even more the return is not
> also detected (!)
>
> I was thining that Gcc was really strict and I choose it because I was
> thinking that, IÂ’m quite disappointed :/
> I’ve tried –Wall –Wextra –pedantic + a ton of other switch with no luck…
>
> Do i miss something ? 
>
> How can I put Gcc in “user is stupid I need to check all” mode?
>
>
> Laurent
>
>   

RE: G++/GCC not detetcing stupid errors in code.

[Laurent Dufrechou]

Arf no... Here is my compile option :)

g++ -march=nocona -Wall -Wfloat-equal -Wshadow -Wpointer-arith -Wcast-align
-Wcast-qual -Wconversion -Wextra -Wreorder -Wold-style-cast -Winit-self
-pedantic -Wreturn-type -Wunreachable-code -D REENTRANT -D_GNU_SOURCE test.c
-o test

> Implicit conversion are a nice thing to have ;-).
Yeah sure , but I prefer the secure way ;)


-----Message d'origine-----
De : gcc-help-owner@gcc.gnu.org [mailto:gcc-help-owner@gcc.gnu.org] De la
part de Sven Eschenberg
Envoyé : jeudi 22 mai 2008 10:57
À : Laurent Dufréchou
Cc : gcc-help@gcc.gnu.org
Objet : Re: G++/GCC not detetcing stupid errors in code.

You are looking for -Wconversion I guess.

Implicit conversion are a nice thing to have ;-).

Regards

-Sven


Laurent Dufréchou schrieb:
> Hi there,
>
> I’m using g++ for the first time via gcc 4.1.2.
> I’ve written a little code (first C++ oriented and because guys with which
I
> work come from C, it is slightly modified to compile with G++ and gcc)
>
>   
>>>>>>>>>>>>>>>>>>>> Code >>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>                                         
>  
> //#include <iostream>
> #include <stdio.h>
>
> //using namespace std;
>
> int main()
>
> {
>
>                 int toto;
>
>                 short tata;
>
>                 toto = 1111118;
>
>                 tata = toto;  //error 1!
>
>                 //cout << toto << "tata:" << tata << endl;
>
>                 printf("%d %i",toto,tata);
>
>                 //return 0; //error 2 because no return
>
> }
>
>   
>
> I i compile with gcc (cc), it detects no int returned error (OK good
point)
> But the downcast (tata = toto) doesn’t raise any warning!
>
> If I use G++ the downcast is not detected and even more the return is not
> also detected (!)
>
> I was thining that Gcc was really strict and I choose it because I was
> thinking that, I’m quite disappointed :/
> I’ve tried –Wall –Wextra –pedantic + a ton of other switch with no luck…
>
> Do i miss something ? 
>
> How can I put Gcc in “user is stupid I need to check all” mode?
>
>
> Laurent
>
>   

Re: G++/GCC not detetcing stupid errors in code.

[Tom Browder]

On Thu, May 22, 2008 at 4:00 AM, Laurent Dufrechou
<laurent.dufrechou@free.fr> wrote:
> Arf no... Here is my compile option :)
>
> g++ -march=nocona -Wall -Wfloat-equal -Wshadow -Wpointer-arith -Wcast-align
> -Wcast-qual -Wconversion -Wextra -Wreorder -Wold-style-cast -Winit-self

The failure to warn about the downcast may be a known problem.  I had
a problem the other way on a later version of gcc, and the gcc-dev
list has some threads about such.

I suggest checking bugzilla and also check at least gcc version 4.3.

-Tom

RE: G++/GCC not detetcing stupid errors in code.

[Laurent Dufréchou]

Yeah compiling with g++-4.3 solved the issue of downcasting. I've now a big
warning :) COOL!
Thanks Tom.
Any idea about the (not so worrying) no return in function requesting
return?
Gcc detect it but not g++. 
Perhaps I miss another keyword???

If you've got no idea, I will fill a bug report :)

Lauennt (code below)

//#include <iostream>
#include <stdio.h>

//using namespace std;

int main()

{

                int toto;

                short tata;

                toto = 1111118;

                tata = toto;  //error 1!

                //cout << toto << "tata:" << tata << endl;

                printf("%d %i",toto,tata);

                //return 0; //error 2 because no return <-----------------

}

-----Message d'origine-----
De : gcc-help-owner@gcc.gnu.org [mailto:gcc-help-owner@gcc.gnu.org] De la
part de Tom Browder
Envoyé : jeudi 22 mai 2008 13:00
À : Laurent Dufrechou
Cc : Sven Eschenberg; Laurent Dufréchou; gcc-help@gcc.gnu.org
Objet : Re: G++/GCC not detetcing stupid errors in code.

On Thu, May 22, 2008 at 4:00 AM, Laurent Dufrechou
<laurent.dufrechou@free.fr> wrote:
> Arf no... Here is my compile option :)
>
> g++ -march=nocona -Wall -Wfloat-equal -Wshadow -Wpointer-arith
-Wcast-align
> -Wcast-qual -Wconversion -Wextra -Wreorder -Wold-style-cast -Winit-self

The failure to warn about the downcast may be a known problem.  I had
a problem the other way on a later version of gcc, and the gcc-dev
list has some threads about such.

I suggest checking bugzilla and also check at least gcc version 4.3.

-Tom

Re: G++/GCC not detetcing stupid errors in code.

[Andrew Haley]

Laurent Dufréchou wrote:
> Yeah compiling with g++-4.3 solved the issue of downcasting. I've now a big
> warning :) COOL!
> Thanks Tom.
> Any idea about the (not so worrying) no return in function requesting
> return?
> Gcc detect it but not g++. 
> Perhaps I miss another keyword???
> 
> If you've got no idea, I will fill a bug report :)

No, it's your knowledge of C++.

ISO/IEC 14882:1998(E)
3.5 Program and linkage

"A return statement in main has the effect of leaving the main
function (destroying any objects with automatic storage duration) and
calling exit with the return value as the argument. If control reaches
the end of main without encountering a return statement, the effect is
that of executing

return 0;"

Note that this is a difference between C and C++: in C you must have
a return value.

Andrew.

Re: G++/GCC not detetcing stupid errors in code.

[Axel Freyn]

Hello Laurent,

On Thu, May 22, 2008 at 01:54:14PM +0200, Laurent Dufréchou wrote:
> [...]
> Any idea about the (not so worrying) no return in function requesting
> return?
> Gcc detect it but not g++. 
> Perhaps I miss another keyword???
> 
> If you've got no idea, I will fill a bug report :)
> [...]
I would expect that this warning does not exist, because omitting the return-statement in
"main" is perfectly valid in C++: (C++-Standard, 3.6.1.3)
"If control reaches the end of main without encountering a return statement, the
effect is that of executing return 0;"

For other functions, g++ also emits a warning if you forget the return
statement.

Axel

RE: G++/GCC not detetcing stupid errors in code.

[Laurent Dufréchou]

>No, it's your knowledge of C++, etc...
>Note that this is a difference between C and C++: in C you must have
>a return value.

Ok! Good, so there is no error in the code due to forced return 0 all is ok
now :)
Will read :
http://www.hep.wisc.edu/~pinghc/isocppstd/basic.html

Thx for the support,
Laurent

-----Message d'origine-----
De : Andrew Haley [mailto:aph@redhat.com] 
Envoyé : jeudi 22 mai 2008 14:09
À : Laurent Dufréchou
Cc : 'Tom Browder'; 'Sven Eschenberg'; gcc-help@gcc.gnu.org
Objet : Re: G++/GCC not detetcing stupid errors in code.

Laurent Dufréchou wrote:
> Yeah compiling with g++-4.3 solved the issue of downcasting. I've now a
big
> warning :) COOL!
> Thanks Tom.
> Any idea about the (not so worrying) no return in function requesting
> return?
> Gcc detect it but not g++. 
> Perhaps I miss another keyword???
> 
> If you've got no idea, I will fill a bug report :)

No, it's your knowledge of C++.

ISO/IEC 14882:1998(E)
3.5 Program and linkage

"A return statement in main has the effect of leaving the main
function (destroying any objects with automatic storage duration) and
calling exit with the return value as the argument. If control reaches
the end of main without encountering a return statement, the effect is
that of executing

return 0;"

Note that this is a difference between C and C++: in C you must have
a return value.

Andrew.

Re: G++/GCC not detetcing stupid errors in code.

[Tom Browder]

On Thu, May 22, 2008 at 6:54 AM, Laurent Dufréchou
<laurent.dufrechou@gmail.com> wrote:
> Yeah compiling with g++-4.3 solved the issue of downcasting. I've now a big
> warning :) COOL!
> Thanks Tom.

You're welcome.  I suggest always staying with the latest releases of
gcc if you can.  It is always being improved and it forces users to
clean up their code and adhere to standards.

-Tom

Re: G++/GCC not detetcing stupid errors in code.

[Christopher Layne]

On Thu, May 22, 2008 at 02:22:49PM +0200, Axel Freyn wrote:
> I would expect that this warning does not exist, because omitting the return-statement in
> "main" is perfectly valid in C++: (C++-Standard, 3.6.1.3)
> "If control reaches the end of main without encountering a return statement, the
> effect is that of executing return 0;"
> 
> For other functions, g++ also emits a warning if you forget the return
> statement.
> 
> Axel

This is something, that while valid by the C++ "Standard", I surely hope
doesn't become common practice.

-cl