RE: False positive warning: missing return statement and switch statements

RE: False positive warning: missing return statement and switch statements

[Steven L. Zook]

What happens when someone calls with:

toInt( (Number)3 );

C/C++ lets you cast any numeric to an enum.

-----Original Message-----
From: gcc-help-owner@gcc.gnu.org [mailto:gcc-help-owner@gcc.gnu.org] On
Behalf Of Frans Englich
Sent: Tuesday, November 15, 2005 2:21 PM
To: gcc-help@gcc.gnu.org
Subject: False positive warning: missing return statement and switch
statements


Hello,

For the code below, when I invoke "g++ -Wall file.cpp" with GCC version
3.3.4, I get:

file.cpp: In function `int toInt(Number)':
file.cpp:20: warning: control reaches end of non-void function

------------------------------------------------------------------------
-
enum Number
{
        Zero,
        One,
        Two
};

int toInt(const Number num)
{
        switch(num)
        {
                case Zero:
                        return 0;
                case One:
                        return 1;
                case Two:
                        return 2;
        }
}

int main()
{
        Number num = Zero;
        return toInt(num);
}
------------------------------------------------------------------------
-

I don't understand why the warning is issued. Isn't it fair to assume
that the argument(const Number num) is correct since it is strongly
typed, by being an enumerator?


Cheers,

		Frans

Re: False positive warning: missing return statement and switch statements

[Frans Englich]

On Tuesday 15 November 2005 22:26, Steven L. Zook wrote:
> What happens when someone calls with:
>
> toInt( (Number)3 );
>
> C/C++ lets you cast any numeric to an enum.

Indeed. I think it's easy to deliberately break code which gcc doesn't warn 
for. If the programmer decides to by pass static type checking and (ab)use an 
enumerator by not using it for what it is but as a plain integer -- then I 
expect it to break(no return value at runtime). After all, if one want 
arbitrary integers: then use an integer, not an enum.

As I see it, the whole point with enumerators is static type checking safety. 
The warning counter-acts this by giving false positives for those who use 
enumerators properly(that is, they don't store arbitrary integers in them).

Obviously, I am looking at this the wrong way -- please explain.


Cheers,

		Frans

Re: False positive warning: missing return statement and switch statements

[John Love-Jensen]

Hi Frans,

> Obviously, I am looking at this the wrong way -- please explain.

Improve your code safety using C++ capabilities:

int toInt(const Number num)
{
  switch(num)
  {
  case Zero:
    return 0;
  case One:
    return 1;
  case Two:
    return 2;
  default:
    throw std::range_error("toInt(): enum Number out of range");
  }
}

Then the warning goes away, and your code is better.

Sincerely,
--Eljay

Re: False positive warning: missing return statement and switch statements

[Frans Englich]

On Tuesday 15 November 2005 22:47, John Love-Jensen wrote:
> Hi Frans,
>
> > Obviously, I am looking at this the wrong way -- please explain.
>
> Improve your code safety using C++ capabilities:
>
> int toInt(const Number num)
> {
>   switch(num)
>   {
>   case Zero:
>     return 0;
>   case One:
>     return 1;
>   case Two:
>     return 2;
>   default:
>     throw std::range_error("toInt(): enum Number out of range");
>   }
> }
>
> Then the warning goes away, and your code is better.

Yupp, that's what I currently do, in my case it unfortunately becomes a 
significant amount of code for dummy tests(because I know I'm behaving).


Cheers,

		Frans

Re: False positive warning: missing return statement and switch statements

[Frans Englich]

On Tuesday 15 November 2005 23:26, Steven L. Zook wrote:
[...]

> I guess that the compiler writers have a very difficult choice in these
> kind of warnings between under and over warning. Their choice may not
> work well for you.

Yes, what a tricky(and flame catching) task of designing warnings. If only it 
was possible to please all.


Cheers,

		Frans

RE: False positive warning: missing return statement and switch statements

[Steven L. Zook]

int toInt(const Number num)
{
        switch(num)
        {
                case Zero:
                        return 0;
                case One:
                        return 1;
                case Two:
                default:
                        return 2;
        }
} 

This approach at least returns something without causing much code
expansion. If you want to trust your code, this approach will silence
the warnings without costing much at run-time.

I guess that the compiler writers have a very difficult choice in these
kind of warnings between under and over warning. Their choice may not
work well for you. 

-----Original Message-----
From: Frans Englich [mailto:frans.englich@telia.com] 
Sent: Tuesday, November 15, 2005 3:01 PM
To: John Love-Jensen
Cc: Steven L. Zook; MSX to GCC
Subject: Re: False positive warning: missing return statement and switch
statements

On Tuesday 15 November 2005 22:47, John Love-Jensen wrote:
> Hi Frans,
>
> > Obviously, I am looking at this the wrong way -- please explain.
>
> Improve your code safety using C++ capabilities:
>
> int toInt(const Number num)
> {
>   switch(num)
>   {
>   case Zero:
>     return 0;
>   case One:
>     return 1;
>   case Two:
>     return 2;
>   default:
>     throw std::range_error("toInt(): enum Number out of range");
>   }
> }
>
> Then the warning goes away, and your code is better.

Yupp, that's what I currently do, in my case it unfortunately becomes a
significant amount of code for dummy tests(because I know I'm behaving).


Cheers,

		Frans

False positive warning: missing return statement and switch statements

[Frans Englich]

Hello,

For the code below, when I invoke "g++ -Wall file.cpp" with GCC version 3.3.4, 
I get:

file.cpp: In function `int toInt(Number)':
file.cpp:20: warning: control reaches end of non-void function

-------------------------------------------------------------------------
enum Number
{
        Zero,
        One,
        Two
};

int toInt(const Number num)
{
        switch(num)
        {
                case Zero:
                        return 0;
                case One:
                        return 1;
                case Two:
                        return 2;
        }
}

int main()
{
        Number num = Zero;
        return toInt(num);
}
-------------------------------------------------------------------------

I don't understand why the warning is issued. Isn't it fair to assume that the 
argument(const Number num) is correct since it is strongly typed, by being an 
enumerator?


Cheers,

		Frans