Re:Cap'n'Proto remote vuln: pointer overflow check optimized...
by johncolanduoni on Friday April 21, 2017 @05:22AM
(#14164114)
Is it that unusual? I've seen undefined behavior be specifically mentioned in both the Java and CLR standards. Rust's (admittedly squishy) "reference" also specifies a lot of behavior as undefined explicitly (along with a general "if you break mandated invariants behavior is undefined").
[ Parent
| on HN ]
Re:Cap'n'Proto remote vuln: pointer overflow check optimized...
by steveklabnik on Friday April 21, 2017 @09:48AM
(#14165548)
In Rust's case, this is limited to unsafe code, at least; safe code should not have UB. (Though, IIRC there is one case right now, but that's a bug) UB is basically "breaking the invariants safe code expects."
Overflow is an example of something that's not UB in Rust, for example. (It's a "program error" and well-defined as two's compliment wrapping.)