3 ms·
> but those failure modes are not expressed by the source code of the running program, they're triggered at a layer of abstraction well below the compiled and r
by Dylan16807 3y ago
> but those failure modes are not expressed by the source code of the running program, they're triggered at a layer of abstraction well below the compiled and running software
Does that make a difference? There's no way to use source code without applying the real world to it.
> is this acceptable? can i assume that calling x will never panic, in practice?
If the probability is that high, then even if x was just { return 123 } you shouldn't assume it will work.
If you replace that code with if (secure_rand_u128() == 0) { panic("boom") }, then it would be safe to assume it won't panic.
- preseinger 3y ago> If the probability is that high, then even if x was just { return 123 } you shouldn't assume it will work. if x -> { return 123 } then callers can 100% assume it will work what is the alternative? how could callers de-risk the case when a solar ray flips a bit to make x not return 123? they can't, because they don't have access to an execution model outside of the semantics of the language, and those language semantics literally guarantee x will return 123 this is literally basic computer science material > if you replace that code with if (secure_rand_u128() == 0) { panic("boom") }, then it would be safe to assume it won't panic. secure_rand_u128 returns, presumably, a random unsigned 128 bit integer unsigned 128 bit integers can be any value between 0 and 340282366920938463463374607431768211455 -- and one of those possible return values is 0 if i call secure_rand_u128, then it absolutely can return 0, and code which assumes that it will never return 0 is, factually, buggy if you want to make the assumption that secure_rand_u128 will never return 0, then you have to justify that assumption in the context of your specific use case -- it absolutely cannot be assumed in the general case without qualifications -- P(x|x>0) != P(0)