How can I cause integer overflows to make an exception in matlab scripts?

I am attempting to write a unit test in MATLAB scripts for a function that is used in an EML block in Simulink. In the EML block when there is an integer overflow I have it configured to throw an exception. In Matlab scripts when there is an integer overflow in the same function, it silently saturates the operation. How do I get the function when called from a Matlab script to behave the same as when called from an EML block? (Note: configuring the EML block to saturate on integer overflows is NOT an option).

I am using R2012b Matlab and Simulink.

[Jan]

This is one of the very rare cases where the tag "matlab code" makes any sense.

Answer 1

[Jan]

You can't.

The silent saturation is a built-in behavior, whcih can be changed only by shadowing the concerned operators:

 File: \@uint8\plus.m

function c = plus(a, b)
c = a + b;

t = double(a) + double(b);
if t > 255
  error('Saturation!');
end

% Or:
% d = c - a;
% if d ~= b, error('Saturation!'); end

But such code needs an exhaustive testing also, because it is too easy to oversee some effects. Creating unit tests for testing unit tests based on shadowed builtin functions is not a clean way to work.

Therefore I suggest to create a new class for integer types with overflow. Implementing the operators in C would allow for a fair speed, but even then detecting overflows is not trivial.

[Jay Gowdy]

At one time, many revisions ago, I remember a parameter setting that made this behavior possible (stopping scripts on integer overflows instead of saturation). It is sad it went away, because it makes developing in MATLAB with the ultimate goal of making EML blocks in Simulink much harder (because you can never know if the results you see in your otherwise EML-compliant MATLAB script are what you are going to see in your equivalent Simulink EML block).