demonstration of a soundness flaw in rmp_serde

raw_unsound.rs

use rmp_serde::Raw;
use serde::{ser, Serialize};
use thiserror::Error;

/// Serializer that serializes strings as a list of char
struct CharListSerializer {
    output: Vec<char>,
}

#[derive(Debug, Error)]
#[error("error during serialization")]
struct SerializeError;

impl ser::Error for SerializeError {
    fn custom<T>(_: T) -> Self
    where
        T: std::fmt::Display,
    {
        Self
    }
}

impl ser::Serializer for &mut CharListSerializer {
    type Ok = ();
    type Error = SerializeError;

    type SerializeSeq = ser::Impossible<(), SerializeError>;
    type SerializeTuple = ser::Impossible<(), SerializeError>;
    type SerializeTupleStruct = ser::Impossible<(), SerializeError>;
    type SerializeTupleVariant = ser::Impossible<(), SerializeError>;
    type SerializeMap = ser::Impossible<(), SerializeError>;
    type SerializeStruct = ser::Impossible<(), SerializeError>;
    type SerializeStructVariant = ser::Impossible<(), SerializeError>;

    fn serialize_bool(self, _: bool) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_i8(self, _: i8) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_i16(self, _: i16) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_i32(self, _: i32) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_i64(self, _: i64) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_u8(self, _: u8) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_u16(self, _: u16) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_u32(self, _: u32) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_u64(self, _: u64) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_f32(self, _: f32) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_f64(self, _: f64) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_char(self, c: char) -> Result<Self::Ok, Self::Error> {
        self.output.push(c);
        Ok(())
    }

    fn serialize_str(self, s: &str) -> Result<Self::Ok, Self::Error> {
        self.output.extend(s.chars());
        Ok(())
    }

    fn serialize_bytes(self, _: &[u8]) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_none(self) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_some<T: ?Sized>(self, _: &T) -> Result<Self::Ok, Self::Error>
    where
        T: serde::Serialize,
    {
        Err(SerializeError)
    }

    fn serialize_unit(self) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_unit_struct(self, _: &'static str) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_unit_variant(
        self,
        _: &'static str,
        _: u32,
        _: &'static str,
    ) -> Result<Self::Ok, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_newtype_struct<T: ?Sized>(
        self,
        _: &'static str,
        _: &T,
    ) -> Result<Self::Ok, Self::Error>
    where
        T: serde::Serialize,
    {
        Err(SerializeError)
    }

    fn serialize_newtype_variant<T: ?Sized>(
        self,
        _: &'static str,
        _: u32,
        _: &'static str,
        _: &T,
    ) -> Result<Self::Ok, Self::Error>
    where
        T: serde::Serialize,
    {
        Err(SerializeError)
    }

    fn serialize_seq(self, _: Option<usize>) -> Result<Self::SerializeSeq, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_tuple(self, _: usize) -> Result<Self::SerializeTuple, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_tuple_struct(
        self,
        _: &'static str,
        _: usize,
    ) -> Result<Self::SerializeTupleStruct, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_tuple_variant(
        self,
        _: &'static str,
        _: u32,
        _: &'static str,
        _: usize,
    ) -> Result<Self::SerializeTupleVariant, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_map(self, _: Option<usize>) -> Result<Self::SerializeMap, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_struct(
        self,
        _: &'static str,
        _: usize,
    ) -> Result<Self::SerializeStruct, Self::Error> {
        Err(SerializeError)
    }

    fn serialize_struct_variant(
        self,
        _: &'static str,
        _: u32,
        _: &'static str,
        _: usize,
    ) -> Result<Self::SerializeStructVariant, Self::Error> {
        Err(SerializeError)
    }
}

fn main() {
    let mut ser = CharListSerializer { output: Vec::new() };
    let data = "ABC";
    data.serialize(&mut ser).unwrap();
    assert_eq!(ser.output, ['A', 'B', 'C']);
    ser.output.clear();

    // An emoji: 😣
    let data = [0xf0, 0x9f, 0x98, 0xa3];
    let data = Raw::from_utf8(data.as_slice().to_owned());
    data.serialize(&mut ser).unwrap();
    assert_eq!(ser.output, ['😣']);
    ser.output.clear();

    // Uh oh
    let data = [0xf0, 0x9f];
    let data = Raw::from_utf8(data.as_slice().to_owned());
    data.serialize(&mut ser).unwrap();
}