Using local parameters by reference

[tristan70]

Hi All,

I am maintaining some code that looks a bit quirky - but seems to behave as expected. I'd rather not change it, as the interface ("second_function_called" in the below example) is used everywhere, but if anyone believes the code looks illegal, please tell. It's certainly not best practice.

The program is single-threaded. I've simplified it for the purpose of analysis:

void first_function_called(int a)
{
  // pass parameter "a" through to another function for read access only

  second_function_called(&a);
}

void second_function_called(int *b)
{
  some_struct s;

  // Get a copy of b and use it for something

  s.some_field = *b;

  // Parameter "b" is never manipulated, passing a copy of "b" would have been fine!!!
}

I guess the question is - is taking the address of a function parameter and then passing it off to another function for read access OK? I believe it probably is, as it should just be taking the address of the parameter off the first functions stack space.

Thankyou.

[brewbuck]

You'd usually see something like this when code has changed -- perhaps the value used to be altered instead of just read.

As you say, I'd keep the API the same to avoid breaking a bunch of code. But I'd also at least change it to a const pointer to enforce the fact that the second_function_called() does not change the value it points to.

[tabstop]

I'm guessing the point is given in your first comment -- second_function_called can change the value of the parameter passed in (I can't say for sure that it does, but I would imagine so) -- the first_function_called thus "protects" the original value by letting the second function change a harmless copy.

[tristan70]

You'd usually see something like this when code has changed -- perhaps the value used to be altered instead of just read.

As you say, I'd keep the API the same to avoid breaking a bunch of code. But I'd also at least change it to a const pointer to enforce the fact that the second_function_called() does not change the value it points to.

Good tip brewbuck. I've already added "const" to quite a few function headers already, this is just another

This code is legit though right? I've never seen it do anything weird - and to the second reply of this post, no, the parameter pointed to by "b" is never modified.

[brewbuck]

This code is legit though right? I've never seen it do anything weird - and to the second reply of this post, no, the parameter pointed to by "b" is never modified.

It's probably not how you'd write it if you wrote it today, but it's legitimate. This is just what happens when code evolves.