Ok, first draft here: https://github.com/alex/rfcs/blob/usize-overflow/text/0000-usize-panic-overflow.md I'd appreciate any and all feedback before I actually submit.
Looks good to me, except the "4 vulnerabitilities seen thus far all use
usize" does not feel like a sufficient rationale. Perhaps it would be worth noting that
usize generally occurs when you're dealing with pointers and in-memory sizes, and does not make much sense otherwise.
If there's more samples elsewhere, I'd love to review them :-)
nothing in rust-fuzz trophy case at least
Nor in the rustsec advisory DB
I'm tempted to make a meme about the situation or something. How often is it that you're trying to secure something, and your problem is that you don't have enough vulnerabilities?
Judging by the track record so far, I'm pretty sure a bunch of them lurk in the stdlib, but my plan to extract them has hit a snag, see "automated fuzzing of stdlib" topic
Well, it's submitted: https://github.com/rust-lang/rfcs/pull/2635 feedback very much appreciated
@Shnatsel it's funny, that's a bit how I feel about RustSec. there were so many type confusion vulnerabilities in Ruby leading to remote code execution, and Rust has "sadly" eliminated many of those, so RustSec doesn't have the constant barrage of those to justify its existence