.... it's a feature from preventing idiot users from using a weak key :P

The Good News is that if that's the only bug, then Intel has done pretty good.

The Bad News is that, in Software Testing, the all zero's case is called a Boundary Condition and is a necessary test in any test effort. If there's one place no bug should be found it would be there. The fact a bug is found there makes me worry what else may have slipped through the cracks.

Robb Topolski

disclosure: I am a former Intel employee and a software testing professional. I hold Intel in high regard but I am not beholden to them. I speak for myself.

Indeed sir, the most cursory and superficial test data should easily have uncovered this "oversight". As an ex-Cobol programmer I recall with delight those fun-filled afternoons when people would ask me to provide them with a file of test data to check the strength and security of their new modules.

Why me? Because I had a knack of breaking things... I always inserted control keys, functions, illogical times, tabs, keyboard commands (like ctrl-v or pgdn), zero length and zero content (escape, delete etc), lines of zeroes, lines of spaces, alt key plus digits, multiple inputs, punctuation (commas, exclamations, pipes etc), wrong character lengths and some interesting entries from the full ASCII character set within the test file. If it passed that lot, it tended to work fairly well - even if the calculation was pants.

Have a jacket to go with your pants.