cin error returning

[boogerlad]

Let's just say that you have

code:

int x = 5; cin >> x; cout << x;

and say you input "derp"
Why does 0 get printed? Since it failed, shouldn't it not modify x at all? Or is it because it fails, its return code is false, which is a zero?

[Insectoid]

C++ won't return an error when this happens. It will just continue running. It appears to not change the RAM if this happens, in which case, it will output the last value of x or in this case, whatever the last program to use this particular bit of ram set it to. A lot of RAM is zeros, in my experience, but if you do this enough you may end up with other values.

[boogerlad]

but wouldn't the last value of x be 5, because the cin failed?

[Insectoid]

Oh, I didn't see that you had that. On my computer, the output is still 5. This isn't standardized behavior. The output may depend on your architecture, OS or compiler.

[boogerlad]

I'm beginning to hate gcc for all these little quirks. I'm going to try with a different compiler, and see what happens.

[Insectoid]

I suggest you keep using gcc. It's the industry standard for linux/mac. These aren't 'quirks'. These are things that just don't matter. Why do you care what happens if you try to store a string into an integer? That should never happen in the first place. Odds are, most other compilers will do the same thing, or something even more silly.

[boogerlad]

I switched to tiny c compiler, and I got a zero instead of a five. Same result as gcc. What compiler/os/cpu are you using? I know it's a miniscule thing, but say that you've been doing multiple operations to a variable, and then if a read fails, you want the variable to maintain its value, not a zero?

[Insectoid]

I'm using gcc on OSX with an x86-64 cpu. Ideally, you'll write your code so that a read cannot fail (for example, read into a char array, check if it's an integer, and if it is, assign it to your int variable. Otherwise throw an error message and ask for input again).

[boogerlad]

well, I do have proper error handling code, to check for ints.

code:

int myNumber = 0;  while (true) {    cout << "Please enter a valid number: ";    getline(cin, input);    stringstream myStream(input);    if (myStream >> myNumber)      break;    cout << "Invalid number, please try again" << endl;  }  cout << "You entered: " << myNumber << endl << endl;

but I'm just curious. Even with proper error handling, the variable will still be set to zero until proper input is given.

[boogerlad]

in msvc++ 2010, the program just crashes lol. Now that I know there is no standard, I'll stop worrying about it. =)

[[Gandalf]]

Behaviour like this is usually called "undefined", and C++ is FULL of situations that may cause undefined behaviour. The key is to avoid those situations.

[OneOffDriveByPoster]

boogerlad @ Wed Jan 04, 2012 10:43 pm wrote:

in msvc++ 2010, the program just crashes lol. Now that I know there is no standard, I'll stop worrying about it. =)

There _is_ a Standard.

C++03 subclause 22.2.2.1.2 [lib.facet.num.get.virtuals] paragraph 1:
If an error occurs, val is unchanged; otherwise it is set to the resulting value.

Which is superceded only recently by C++2011 where the expected result is indeed 0.

[Velocity]

you cant return the same value that you recieved without checking a new pub

[QuantumPhysics]

Why would you need to use string handling anyways you could just use the basic iostream... there is no need to call #include <string> it is useless unless your trying to figure out some sort of cipher encryption or decryption

[mirhagk]

@QuantumPhysics that statement isn't even related to what was being talked about (noone mentioned strings) and including <string> gives you a string library that many API's expect (and is usually a lot easier to work with). Plus your necro-ing a topic that is from January, so the original posters likely won't ever see what your saying (I know velocity won't, he's banned I'm pretty sure)