Division overflow exception: bug or feature?

Reinforcements are coming:

There were some comments to the previous posts on division overflow suggesting that throwing an exception is not appropriate.

E.g. fatoff thinks that throwing the exception in this case is not good.

Also, there was an idea that the right code for the job would look like:
return (int) ((long) x / (long) y);
which does not crash on signed int overflow. Unfortunately this code produces an incorrect result, which will almost certainly lead to a hard to find bug once the result is used in the future.

I believe "Fail-fast" is a good thing. The actors here are the C compiler, the CPU, and the programmer. Who should enable fail fast?

The C compiler cannot do it because behavior on signed integer overflow is undefined by the most recent C standards.

The programmer should be aware about the problem but we cannot count on it.

So throwing the exception which facilitates fail fast is a useful CPU feature. Besides, this X86 feature is well documented in the Intel manual.