detect underflow/overflow error at runtime?

Question

Out of curiosity, is there any way to detect underflow/overflow errors for numbers at runtime? if no, why not? I know it can be the expected behavior but it would still be helpful I'm using Visual Studio 2010 if that changes anything

edit: for instance:

unsigned int a= 2;
unsigned int b= 3;
a -=b; //<- underflow

This has already been asked here: http://stackoverflow.com/questions/2399269/checking-for-underflow-overflow-in-c — dario_ramos, Jan 04 '12 at 17:51
@dario_ramos : I was thinking more of a "feature" that would not need specific checks everytime — lezebulon, Jan 04 '12 at 17:53
So... `if (b <= a) a -= b; else throw some_error();`? Unsigned integer arithmetic is well-defined by the standard, and never an "error". The question would be more interesting for *signed* integers. — Kerrek SB, Jan 04 '12 at 17:54
@Kerrek : even if it's not UB in some cases, I'm sure that unwanted underflow/overflow causes a lot of bugs and it'd be helpful to have some automatic detection — lezebulon, Jan 04 '12 at 18:00
@lezebulon: sometimes yes, but it'd make the program go a _lot_ slower. — Mooing Duck, Jan 04 '12 at 18:03
For VC++ specifically, see MS' [`SafeInt` library](http://msdn.microsoft.com/en-us/library/dd570023.aspx). — ildjarn, Jan 04 '12 at 18:10
I know I've seen other libraries for this, but I can't find the term for it, so I can't find the libraries :( — Mooing Duck, Jan 04 '12 at 18:12
[A clamped integer](http://www.google.com/search?q=C%2B%2B+clamped+integer) — Mooing Duck, Jan 04 '12 at 18:18

score 4 · Accepted Answer · answered Jan 04 '12 at 18:01

C++ is a relatively lightweight language and as such doesn't provide any sort of automatic runtime checking for overflow/"underflow". Your code should generally be written in such a way that such things won't happen, and in cases where you're dealing with external input you'll need to guard the inputs with code specifically designed for the particular case you're protecting.

score 2 · Answer 2 · answered Jan 04 '12 at 18:28

2

Visual Studio includes an implementation of the SafeInt class which by default will raise an exception when an overflow/underflow occurs.

answered Jan 04 '12 at 18:28

Bukes

3,668
1
18
20

1

I'm doing a little research on this, and thought I'd mention that besides having a variant that ships in Visual Studio, the library is hosted on codeplex and released under the ostensibly-permissive [Microsoft Public License](http://safeint.codeplex.com/license), and works in recent versions of gcc as well. – HostileFork says dont trust SE Jun 17 '12 at 04:58

detect underflow/overflow error at runtime?

2 Answers2