Convert crash on binop to controlled error

[septract]

Partially addresses #233

[septract]

@cp526 I guessed at an error message, but perhaps you can propose something better?

[cp526]

This should probably mention that this is unsupported on anything other than unsigned integers.

[cp526]

... Although I think these might really be proper assertion failures.

[septract]

What does it mean for this to be a proper assertion failure? I would assume these should only occur in anomalous states, ie states that should be prevented by some other CN code

[cp526]

I suspect we're in that kind of situation here: something in CN+Cerberus's earlier logic should prevent an application of M_PEbounded_binop (M_Bound_Wrap, ...) when the C-type is a signed integer (wrapping behaviour typically only happens for unsigned types).

Similarly, the M_PEbounded_binop below should only happen on integer types.

[septract]

Huh, okay - so is the conclusion that the crash in #233 is actually a bug in CN, rather than an unsupported feature? That seems more difficult to resolve, & probably needs you to take a deeper look @cp526

[cp526]

Yes, I think so; I see no reason why CN shouldn't handle the example from #233 .

[septract]

Okay, I'm closing this PR - looks like someone should look at the root cause of #233 (I'll take a look but I suspect it's too deep in the stack for my level of understanding)