← Back to context

Comment by oneeyedpigeon

2 days ago

Which of the several flaws in this bug chain would have been materially different with a language other than C?

49 comments

oneeyedpigeon

Reply
nh2  2 days ago

Any use of sum types would have prevented this immediately.

The whole thing exists because APIs that return a NULL value that "nobody checks", plus a string that you shouldn't use if you actually did check the null value.

In languages with sum types, this cannot happen.

Either you get your hash, or no result at all. You couldn't get eg NULL and *.

  • jppittma  2 days ago

    C has unions. C has sum types. People have been doing tagged unions in C for longer than rust has been a language

    • creata  2 days ago

      It is extremely unidiomatic to use tagged unions in C for error handling. It's definitely possible, but awkward, and the compiler doesn't check that you checked the tag.

    • K0nserv  2 days ago

      I don't think I've ever seen a C function that uses unions for this purpose though?

      I can think of two common patterns that I've seen frequently:

      * Return NULL for errors (sometimes with a global errorno() function).

      * Return an integer error with any return values as out parameters.

      So if it is possible to improve, why doesn't anyone do it?

      1 reply →

    • aliceryhl  2 days ago

      Yet the authors of crypt did not use them. The C language is not designed in a way to encourage people to write code that correctly uses tagged unions (or nullable pointers for that matter).

      5 replies →

    • zarzavat  1 day ago

      That's a universe apart from real sum types with exhaustiveness checking and pattern matching.

  • xorcist  2 days ago

    This argument is weak. C, and specifically libc, has errno. The error code is there. Not in band.

    But there has to be something in the payload field when it is not applicable. A null is perfectly fine.

    The argument for changing that to "*", if I understand the above linked discussion correctly, was that some users of this api threw away errors and a dummy value would be harder to ignore. Maybe one could argue that these behaviours are less common in Rust or Go, but not because of a technical difference.

    The world is not perfect and sometimes you really need to work around bad users. Even in Rust or Go. I am not convinced this was such a case. If you are convinced it was it would still have been better to select a less trivial dummy value (such as "*:ERROR:*").

    • forresthopkinsa  2 days ago

      > But there has to be something in the payload field when it is not applicable

      Not in languages with ergonomic sum types. A `Result` or `Either` will have no payload field on the error type. The compiler won't let you dereference it.

    • lemonwaterlime  1 day ago

      In Haskell, one would return a Maybe or an Either here.

          Maybe
    ——————
    Just hash
    Nothing

    Either
    ——————————
    Left err
    Right hash

      The program would be designed around not allowing this failure mode to even be possible.

    • aidenn0  2 days ago

      This is wrong. errno is not to be used to detect errors, only to get the cause of the error once an error has been detected.

  • dolmen  2 days ago

    Nope.

    Because in fact all the changes were about changing the function signatures. The function signature is enforced by the linker. The semantic of strings that go through isn't.

    Returning 2 values is possible in C (via pointers input parameters). The main issue here is that libc is a strong API contract based on C function signatures, that nobody can change, only augment with new symbols. Such strong API compatibility constraints happen in any programming language, and workaround to fit new behaviours in old APIs happens everywhere.

    • K0nserv  2 days ago

      I don't think nh2 is saying that this is how the "issue" should've been solved here, but that it's a fatal flaw of C itself (yes I know, unions, but no one uses them).

      But also, its not like there were many alternative languages when the POSIX standards were designed so it's fair that it works like this.

  • aidenn0  2 days ago

    Even C has a better type system than C: All modern C compilers support an extension that lets you flag a function's return-value as requiring a null check.

muvlon  2 days ago

C's lack of all of

- sum/union types

- multiple return values/tuple types

- exceptions

means that most ways of returning an error from a function are not available in C. Because of this, many C APIs, including `crypt` in this case, resort to in-band error signalling by using a sentinel value (like NULL, a negative integer, "*" etc.).

Most other languages lack one or two of the above, but very few lack all three, and so most have developed a different culture around error reporting (Result in Rust, multiple return values in Go, exceptions in Java, Python etc.).

  • tialaramex  2 days ago

    One of the small but important significances of Rust is that it very often gets to have the same machine code as a correct "in-band error signalling" strategy from C but with source code that the maintainers can actually understand.

    And really often when your "in-band error signalling" in Rust is more expensive you realise (perhaps to your horror) that your C wouldn't be correct, so you just dodged a bullet. For example Option<OwnedFd> is, like OwnedFd, a 32-bit value, Unix has always reserved -1 to mean invalid, so Rust can treat that as a niche. But the Windows equivalent Option<OwnedHandle> can't do this because, well, cue Raymond Chen: https://devblogs.microsoft.com/oldnewthing/20040302-00/?p=40...

    • creata  2 days ago

      > One of the small but important significances of Rust is that it very often gets to have the same machine code as a correct "in-band error signalling" strategy from C but with source code that the maintainers can actually understand.

      But aside from NonZero- and NonNull-based layout optimizations, I don't think Rust is doing anything special to obtain that machine code - it's just relying on LLVM to optimize things.

      1 reply →

  • Gibbon1  1 day ago

    I remember mucking with returning multiple values as anonymous structs. It was tantalizing close but no banana.

    I think with C23 you got half a banana now. Still can't do it because C23 didn't adopt proposals to allow returns of untagged/anonymous structs.

    If you could then you could write

      auto rtn = Foo();
  if(rtn.err) 
    printf("err=%i", rtn.err);
  else
    printf("Foo=%i\n", rtn.foo);

  • jppittma  2 days ago

    C has sum types/unions? What are you talking about? You don't need multiple return. You can use a struct and it's the same thing.

    • adgjlsfhk1  2 days ago

      not in a way that forces the user to actually handle them properly. C is a weakly typed language and as such is pretty badly suited to systems programming.

      5 replies →

    • muvlon  2 days ago

      C unions are sum types if you just "draw the rest of the fucking owl". C has sum types the way x86 assembly has function calls. You see, there are "call" and "ret" instructions. Just figure out calling convention and the stack yourself, it's easy. And yes I can stick the tag next to the union myself in C and write a bunch of functions that hopefully deal with it correctly. I can even write some macros that look like pattern matching and maybe convince myself that it's a language feature, but it just isn't.

      And it's very evident why the difference matters. People gravitate to what's readily available and ergonomic. That's why every C API under the sun (most definitely including libc itself) does janky stuff like returning -1 or "*" on error instead of building their own bespoke Result-style abstraction. A programming language is not just the things that are in some way possible within it, but more importantly the things that are easy in it.

      5 replies →

K0nserv  2 days ago

For the other part of the issue (the truncation in SHA512), in Rust you could've had:

    struct Input<'a>(&'a [u8]);
    

    impl<'a> Input<'a> {
      fn new(input: &'a [u8]) -> Result<Self, Error> {
        if input.len() >= 256 {
           return Err(Errro...);
        }
        
        Ok(Self(input))
      }
    }

Now the truncation is a terrible idea in the first place, but at least with this it would have been entirely obvious to the caller that it was there.

dmantis  2 days ago

Wouldn't Rust's Result or Go's error checking prevent that?

While not mandatory, a person would probably use it instead of the * hack.

Simply because the calling side wouldn't have an option to forget the check like with the NULL case

  • ninkendo  2 days ago

    Rust and go still make it pretty easy to not check results if you don’t care about the success value (lots of code just does `let _ = foo()`), but if you do want the value (like with a hashing function), Rust makes it really hard to not check the error too. There’s no syntax to say “assume this worked and give me the success value, including whatever garbage is there if it’s not set”:

    - result.unwrap() will panic if it failed

    - result.ok() throws away the error but gives you a None value instead (still can’t use it as if it’s a success)

    - matching on the Ok variant and doing nothing on the Err variant, you only get to use the success value inside the Ok arm

    Etc etc.

    The problem is when the compiler says “NULL is a valid instance of T”, which go lets you do, and rust doesn’t (without using the unsafe keyword and intentionally casting a null pointer, which you don’t do by accident.) Go lets you write code like `var hash string; hash, _ = hashfn()` and it compiles and now you have null instead of a valid hash.

    • 65a  1 day ago

      Minor nit, strings are never nil in Go, so that would be "" instead. Usually nil is only a valid instance of pointer-to-T. The point stands though, a common pattern is "defer foo.Close()", the Close() signature almost always returns an error and almost no one checks it...which is probably fine, but not automatically fine.

    • tialaramex  2 days ago

      > NULL is a valid instance of T

      This by the way is Tony Hoare's famous "Billion Dollar Mistake"

    • uecker  2 days ago

      A NULL pointer for error would be a safe interface where you get a trap when there is no value.