3 ms·
One of the very particular "undebuggable" issues (safe) Rust solves is data races. Experience tells us that humans can't successfully reason about non-trivial
by tialaramex 5y ago
One of the very particular "undebuggable" issues (safe) Rust solves is data races.
Experience tells us that humans can't successfully reason about non-trivial concurrent programs unless they exhibit Sequential Consistency. In Rust you're promised this is what you get. Maybe what you wrote is stupid and wrong, but it has Sequential Consistency. "Oh," you exclaim during debugging, "A might happened before B and then we're in a pretty pickle" - you found the bug, now you just need to fix it. This promise is delivered by never allowing code to have references to things some other code might change, thus eliminating data races.
In most other languages that offer concurrency this promise only applies if you wrote a program with no data races and the responsibility to ensure that is with you, so you can accidentally write programs that don't have Sequential Consistency and thus... "Wait, so, A happened before B and B happened before A? Huh? I don't even understand the bug".
- pjmlp 5y agoOnly for data races via threads on the same process, it does nothing to prevent data races via shared memory using IPC mechanisms across processes.
- tialaramex 5y agoHow so? Obviously you can build an unsafe IPC mechanism with concurrent access, label it "safe" when it isn't, and then say "Look at this horrible mess, I blame Rust" but it seems like it'd be faster to just implement std::ops::Index unsafely and then blame Rust because thing[len+1] blew up even though Rust has "memory safety". Now I'm going to write an amusing aside. One way you could get into this trouble is if your hardware allows arbitrary foreign memory writes as "IPC". The BBC microcomputer allowed this over the network! You could send a bunch of bytes over Econet (a 1980s network from Acorn Computers available for the BBC, Electron and Archimedes computers), addressed to another BBC micro on the network, asking they be written to a RAM buffer, and the remote hardware would do so. We used this to run a Multi-user Dungeon at school in about 1989 or so. A "server" ran the actual MUD software, and individual users signed in from a computer on the network around the school to play, when they typed a command their command buffer was transmitted over Econet, and then the server wrote remotely to their display RAM to show the result on their screen. Fortunately your computer is not a BBC Microcomputer, and foreign processes do not (on the whole) get to scribble on your program's memory. So this should not be a problem unless you specifically make this unsafe decision in your Rust program.
- pjmlp 5y agoBecause as explained on rustnomicon, there are many ways to access data in parallel ways, and the language only prevents a tiny subset of those cases. However almost every time data races and Rust come together in the same sentence it is as if it would prevent all use cases.
- tialaramex 5y agoBut (unless you've got an excerpt that says otherwise) the Rustonomicon is about unsafe Rust. And I was explaining that safe Rust has data race freedom. The Rustonomicon is not warning you about scary hidden problems in safe Rust, it's warning you about scary problems you need to care about when writing unsafe Rust, so that your unsafe Rust has appropriate safety rails before anybody else touches it. // Safety: Can't touch this while anybody else might write This reminds me of the mutex thing. Look at C++ std::mutex. You could implement exactly that in Rust. But, that's not what std::sync::Mutex is at all. Because if you implemented it in Rust, C++ std::mutex is either useless or unsafe and clearly we'd prefer neither. But do you have some examples of Rust shared memory IPC that you believe are unsafe? It might be instructive to either show why they're actually safe after all or, alternatively, go add the unsafety explanations and work out what a safe wrapper would look like.
- pjmlp 5y agoYou as a user of a crate deemed safe, that underneath uses shmem, mmap, or a database, written without taking the proper care to prevent other processes to change exactly the same underlying data segment, written in what knows what, are in for a surprise and long debugging sessions. The crate public API surface is safe after all, and unless the user has experience in distributed systems, the answer won't come right away.
- tialaramex 5y agoBut this hypothetical "written without taking the proper care" code is buggy. Like I said it's just the same situation as a bad implementation of Index but more convoluted. Rust's standard library takes this very seriously. In many languages if I try to sort() things which refuse to abide by common sense rules like "Having a consistent sort order" the algorithm used may blow up arbitrarily. But Rust's sort() is robust against that. You may create an infinite loop (legal in Rust, causes Undefined Behaviour in C++) and the result of sorting things without a meaningful ordering is unlikely to be helpful if it does finish, but it's guaranteed to be safe, you won't get Undefined Behaviour.