Question regarding functions returning reference

[zebulon]

Hi,

I am aware that when a function returns a reference, one should never
return a local variable to that function, since that variable will be
removed from the stack and the reference will be dangling.

Therefore, this is bad:
double& GetWeeklyHours()
{
double h =
46.50;
return h;
}

However, why this example (found in http://www.functionx.com/cpp/example...nreference.htm)
seems valid:
double& GetWeeklyHours()
{
double h = 46.50;
double &hours = h;
return hours;
}

Where is the reference &hours created ? Shouldn't it be destroyed when
the function exists too ? Is that ok ? Is that even elegant ?

Thanks for your opinions.

[asm23]

zebulon wrote:

Hi,

I am aware that when a function returns a reference, one should never
return a local variable to that function, since that variable will be
removed from the stack and the reference will be dangling.

Therefore, this is bad:
double& GetWeeklyHours()
{
double h =
46.50;
return h;
}

However, why this example (found in http://www.functionx.com/cpp/example...nreference.htm)
seems valid:
double& GetWeeklyHours()
{
double h = 46.50;
double &hours = h;
return hours;
}

Yes, I test it Using two different compilers, all compile with NO errors
or warnings.

But it seems the function return a reference to a local variable.
I have the same question with you.
Someone can explain it. Thanks.

>
Where is the reference &hours created ? Shouldn't it be destroyed when
the function exists too ? Is that ok ? Is that even elegant ?

Thanks for your opinions.

[zebulon]

On Sep 3, 3:04*pm, asm23 <[e-mail]:

Yes, I test it Using two different compilers, all compile with NO errors
or warnings.

gcc using -Wall returns a warning for the first example, but not the
second one. This is on a Unix machine.

[asm23]

zebulon wrote:

On Sep 3, 3:04 pm, asm23 <[e-mail]:

>Yes, I test it Using two different compilers, all compile with NO errors
or warnings.

gcc using -Wall returns a warning for the first example, but not the
second one. This is on a Unix machine.

I compiled in Visual C++ 6.0 and Intel C++ 9.1, with the default option,
there is no errors or warnings.

[Marco Manfredini]

zebulon wrote:

However, why this example (found in
http://www.functionx.com/cpp/example...nreference.htm) seems
valid: double& GetWeeklyHours()
{
double h = 46.50;
double &hours = h;
return hours;
}

Where is the reference &hours created ? Shouldn't it be destroyed when
the function exists too ? Is that ok ? Is that even elegant ?

Please delete your bookmark to this site immediately: This is complete
bullshit written by an idiot. You are still returning a reference to a
value that gets destroyed on function exit.

[zebulon]

On Sep 3, 3:23*pm, Marco Manfredini <[e-mail]:

Please delete your bookmark to this site immediately: This is complete
bullshit written by an idiot. You are still returning a reference to a
value that gets destroyed on function exit.

Hi Marco,

Thanks for the answer and this actually confirms what I was
suspecting. Unfortunately, this page is ranking second in Google when
querying for "c++ function returning reference"... Aside, g++ does not
warn about the issue in the second example, but the first one only
(using -Wall), maybe a limitation in the warning detection system ?

Zebulon

[red floyd]

On Sep 3, 7:33*am, zebulon <[e-mail]:

On Sep 3, 3:23*pm, Marco Manfredini <[e-mail]:

Please delete your bookmark to this site immediately: This is complete
bullshit written by an idiot. You are still returning a reference to a
value that gets destroyed on function exit.

Hi Marco,

Thanks for the answer and this actually confirms what I was
suspecting. Unfortunately, this page is ranking second in Google when
querying for "c++ function returning reference"... Aside, g++ does not
warn about the issue in the second example, but the first one only
(using -Wall), maybe a limitation in the warning detection system ?

Because it's syntactically correct. The compiler is not required to
issue a diagnostic.
What you have is UB (undefined behavior).

[zebulon]

On Sep 3, 4:11*pm, red floyd <[e-mail]:

Because it's syntactically correct. *The compiler is not required to
issue a diagnostic.
What you have is UB (undefined behavior).

Yes, but that would be the case for the first case too. And there g++
catches the problem. Actually -Wall does not only catch syntax errors,
but also programming errors (as they are warnings, not errors).

Zebulon

[Christian Hackl]

zebulon wrote:

Thanks for the answer and this actually confirms what I was
suspecting. Unfortunately, this page is ranking second in Google when
querying for "c++ function returning reference"... Aside, g++ does not
warn about the issue in the second example, but the first one only
(using -Wall), maybe a limitation in the warning detection system ?

Yes, it is a limitation. Compilers cannot detect all occurrences of
undefined behaviour, nor are they required to by the C++ standard.

You can't count on getting a warning even in blatant cases such as this:

int main()
{
int array[3];
array[5] = 0;
}
--
Christian Hackl

[zebulon]

On Sep 3, 4:51*pm, Christian Hackl <[e-mail]:

Yes, it is a limitation. Compilers cannot detect all occurrences of
undefined behaviour, nor are they required to by the C++ standard.

You can't count on getting a warning even in blatant cases such as this:

Indeed. Thanks for the info.

Zebulon