// Copyright (c) The Libra Core Contributors // SPDX-License-Identifier: Apache-2.0 // For some reason deriving `Arbitrary` results in clippy firing a `unit_arg` violation #![allow(clippy::unit_arg)] use std::{ collections::{BTreeMap, BTreeSet}, fmt, }; use proptest::prelude::*; use proptest_derive::Arbitrary; use serde::{de::DeserializeOwned, Deserialize, Serialize}; use libra_canonical_serialization::{ from_bytes, serialized_size, to_bytes, Error, MAX_CONTAINER_DEPTH, MAX_SEQUENCE_LENGTH, }; fn is_same(t: T) where T: Serialize + DeserializeOwned + fmt::Debug + PartialEq, { let bytes = to_bytes(&t).unwrap(); let s: T = from_bytes(&bytes).unwrap(); assert_eq!(t, s); assert_eq!(bytes.len(), serialized_size(&t).unwrap()); } // TODO deriving `Arbitrary` is currently broken for enum types // Once AltSysrq/proptest#163 is merged we can use `Arbitrary` again. #[derive(Debug, Deserialize, Serialize, PartialEq)] enum E { Unit, Newtype(u16), Tuple(u16, u16), Struct { a: u32 }, } #[test] fn test_enum() { let u = E::Unit; let expected = vec![0]; assert_eq!(to_bytes(&u).unwrap(), expected); is_same(u); let n = E::Newtype(1); let expected = vec![1, 1, 0]; assert_eq!(to_bytes(&n).unwrap(), expected); is_same(n); let t = E::Tuple(1, 2); let expected = vec![2, 1, 0, 2, 0]; assert_eq!(to_bytes(&t).unwrap(), expected); is_same(t); let s = E::Struct { a: 1 }; let expected = vec![3, 1, 0, 0, 0]; assert_eq!(to_bytes(&s).unwrap(), expected); is_same(s); } #[derive(Arbitrary, Debug, Deserialize, Serialize, PartialEq)] struct S { int: u16, option: Option, seq: Vec, boolean: bool, } proptest! { #[test] fn proptest_bool(v in any::()) { assert_eq!(to_bytes(&v)?, vec![u8::from(v)]); is_same(v); } #[test] fn proptest_i8(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_i16(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_i32(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_i64(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_i128(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_u8(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_u16(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_u32(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_u64(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_u128(v in any::()) { assert_eq!(to_bytes(&v)?, v.to_le_bytes()); is_same(v); } #[test] fn proptest_string(v in any::()) { let mut expected = Vec::with_capacity(v.len() + 4); // Larger lengths have more complex uleb128 encodings. prop_assume!(v.len() < 128); expected.extend_from_slice(&(v.len() as u8).to_le_bytes()); expected.extend_from_slice(v.as_bytes()); assert_eq!(to_bytes(&v)?, expected); is_same(v); } #[test] fn proptest_vec(v in any::>()) { let mut expected = Vec::with_capacity(v.len() + 4); // Larger lengths have more complex uleb128 encodings. prop_assume!(v.len() < 128); expected.extend_from_slice(&(v.len() as u8).to_le_bytes()); expected.extend_from_slice(&v); assert_eq!(to_bytes(&v)?, expected); is_same(v); } #[test] fn proptest_option(v in any::>()) { let expected = v.map(|v| vec![1, v]).unwrap_or_else(|| vec![0]); assert_eq!(to_bytes(&v)?, expected); is_same(v); } #[test] fn proptest_btreemap(v in any::, Vec>>()) { is_same(v); } #[test] fn proptest_tuple2(v in any::<(i16, String)>()) { is_same(v); } #[test] fn proptest_tuple3(v in any::<(bool, u32, String)>()) { is_same(v); } #[test] fn proptest_tuple4(v in any::<(bool, u32, Option)>()) { is_same(v); } #[test] fn proptest_tuple_strings(v in any::<(String, String, String)>()) { is_same(v); } #[test] fn proptest_lexicographic_order(v in any::, Vec>>()) { let bytes = to_bytes(&v).unwrap(); // This test assumes small maps and small vectors. // This is what proptest always generates in practice but we will make // the assumptions explicit anyway. prop_assume!(v.len() < 128); let m : BTreeMap, Vec> = v.iter().filter_map(|(k, v)| { if k.len() >= 128 || v.len() >= 128 { return None; } let mut k_bytes = Vec::with_capacity(k.len() + 4); k_bytes.extend_from_slice(&(k.len() as u8).to_le_bytes()); k_bytes.extend(k.iter()); let mut v_bytes = Vec::with_capacity(v.len() + 4); v_bytes.extend_from_slice(&(v.len() as u8).to_le_bytes()); v_bytes.extend(v.iter()); Some((k_bytes, v_bytes)) }) .collect(); prop_assume!(v.len() == m.len()); let mut expected = Vec::with_capacity(bytes.len()); expected.extend_from_slice(&(m.len() as u8).to_le_bytes()); for (key, value) in m { expected.extend(key.iter()); expected.extend(value.iter()); } assert_eq!(expected, bytes); } #[test] fn proptest_box(v in any::>()) { is_same(v); } #[test] fn proptest_struct(v in any::()) { is_same(v); } #[test] fn proptest_addr(v in any::()) { is_same(v); } #[test] fn proptest_bar(v in any::()) { is_same(v); } #[test] fn proptest_foo(v in any::()) { is_same(v); } } #[test] fn invalid_utf8() { let invalid_utf8 = vec![1, 0xFF]; assert_eq!(from_bytes::(&invalid_utf8), Err(Error::Utf8)); } #[test] fn uleb_encoding_and_variant() { #[derive(Serialize, Deserialize, Debug, PartialEq)] enum Test { One, Two, }; let valid_variant = vec![1]; from_bytes::(&valid_variant).unwrap(); let invalid_variant = vec![5]; // Error comes from serde assert_eq!( from_bytes::(&invalid_variant), Err(Error::Custom( "invalid value: integer `5`, expected variant index 0 <= i < 2".into() )) ); let invalid_bytes = vec![0x80, 0x80, 0x80, 0x80]; // Error is due to EOF. assert_eq!(from_bytes::(&invalid_bytes), Err(Error::Eof)); let invalid_uleb = vec![0x80, 0x80, 0x80, 0x80, 0x80]; // Error comes from uleb decoder because u32 are never that long. assert_eq!( from_bytes::(&invalid_uleb), Err(Error::IntegerOverflowDuringUleb128Decoding) ); let invalid_uleb = vec![0x80, 0x80, 0x80, 0x80, 0x1f]; // Error comes from uleb decoder because we are truncating a larger integer into u32. assert_eq!( from_bytes::(&invalid_uleb), Err(Error::IntegerOverflowDuringUleb128Decoding) ); let invalid_uleb = vec![0x80, 0x80, 0x80, 0x80, 0x0f]; // Error comes from Serde because ULEB integer is valid. assert_eq!( from_bytes::(&invalid_uleb), Err(Error::Custom( "invalid value: integer `4026531840`, expected variant index 0 <= i < 2".into() )) ); let invalid_uleb = vec![0x80, 0x80, 0x80, 0x00]; // Uleb decoder must reject non-canonical forms. assert_eq!( from_bytes::(&invalid_uleb), Err(Error::NonCanonicalUleb128Encoding) ); } #[test] fn invalid_option() { let invalid_option = vec![5, 0]; assert_eq!( from_bytes::>(&invalid_option), Err(Error::ExpectedOption) ); } #[test] fn invalid_bool() { let invalid_bool = vec![9]; assert_eq!( from_bytes::(&invalid_bool), Err(Error::ExpectedBoolean) ); } #[test] fn sequence_too_long() { let seq = vec![0; MAX_SEQUENCE_LENGTH + 1]; match to_bytes(&seq).unwrap_err() { Error::ExceededMaxLen(len) => assert_eq!(len, MAX_SEQUENCE_LENGTH + 1), _ => panic!(), } } #[test] fn variable_lengths() { assert_eq!(to_bytes(&vec![(); 1]).unwrap(), vec![0x01]); assert_eq!(to_bytes(&vec![(); 128]).unwrap(), vec![0x80, 0x01]); assert_eq!(to_bytes(&vec![(); 255]).unwrap(), vec![0xff, 0x01]); assert_eq!( to_bytes(&vec![(); 786_432]).unwrap(), vec![0x80, 0x80, 0x30] ); } #[test] fn sequence_not_long_enough() { let seq = vec![5, 1, 2, 3, 4]; // Missing 5th element assert_eq!(from_bytes::>(&seq), Err(Error::Eof)); } #[test] fn map_not_canonical() { let mut map = BTreeMap::new(); map.insert(4u8, ()); map.insert(5u8, ()); let seq = vec![2, 4, 5]; assert_eq!(from_bytes::>(&seq), Ok(map)); // Make sure out-of-order keys are rejected. let seq = vec![2, 5, 4]; assert_eq!( from_bytes::>(&seq), Err(Error::NonCanonicalMap) ); // Make sure duplicate keys are rejected. let seq = vec![2, 5, 5]; assert_eq!( from_bytes::>(&seq), Err(Error::NonCanonicalMap) ); } #[test] fn by_default_btreesets_are_serialized_as_sequences() { // See https://docs.serde.rs/src/serde/de/impls.rs.html // This is a big caveat for us, but luckily, generate-format will track this in the YAML output. let mut set = BTreeSet::new(); set.insert(4u8); set.insert(5u8); let seq = vec![2, 4, 5]; assert_eq!(from_bytes::>(&seq), Ok(set.clone())); let seq = vec![2, 5, 4]; assert_eq!(from_bytes::>(&seq), Ok(set.clone())); // Duplicate keys are just ok. let seq = vec![3, 5, 5, 4]; assert_eq!(from_bytes::>(&seq), Ok(set)); } #[test] fn leftover_bytes() { let seq = vec![5, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]; // 5 extra elements assert_eq!(from_bytes::>(&seq), Err(Error::RemainingInput)); } #[test] fn test_f32() { assert!(to_bytes(&1.0f32).is_err()); } #[test] fn test_f64() { assert!(to_bytes(&42.0f64).is_err()); } #[test] fn test_char() { assert!(to_bytes(&'a').is_err()); } #[test] fn zero_copy_parse() { #[derive(Serialize, Deserialize, Eq, PartialEq, Debug)] struct Foo<'a> { borrowed_str: &'a str, borrowed_bytes: &'a [u8], } let f = Foo { borrowed_str: "hi", borrowed_bytes: &[0, 1, 2, 3], }; { let expected = vec![2, b'h', b'i', 4, 0, 1, 2, 3]; let encoded = to_bytes(&f).unwrap(); assert_eq!(expected, encoded); let out: Foo = from_bytes(&encoded[..]).unwrap(); assert_eq!(out, f); } } #[test] fn cow() { use std::borrow::Cow; let large_object = vec![1u32, 2, 3, 4, 5, 6]; let mut large_map = BTreeMap::new(); large_map.insert(1, 2); #[derive(Serialize, Deserialize, Debug)] enum Message<'a> { M1(Cow<'a, Vec>), M2(Cow<'a, BTreeMap>), } // M1 { let serialized = to_bytes(&Message::M1(Cow::Borrowed(&large_object))).unwrap(); let deserialized: Message<'static> = from_bytes(&serialized).unwrap(); match deserialized { Message::M1(b) => assert_eq!(b.into_owned(), large_object), _ => panic!(), } } // M2 { let serialized = to_bytes(&Message::M2(Cow::Borrowed(&large_map))).unwrap(); let deserialized: Message<'static> = from_bytes(&serialized).unwrap(); match deserialized { Message::M2(b) => assert_eq!(b.into_owned(), large_map), _ => panic!(), } } } #[test] fn strbox() { use std::borrow::Cow; let strx: &'static str = "hello world"; let serialized = to_bytes(&Cow::Borrowed(strx)).unwrap(); let deserialized: Cow<'static, String> = from_bytes(&serialized).unwrap(); let stringx: String = deserialized.into_owned(); assert_eq!(strx, stringx); } #[test] fn slicebox() { use std::borrow::Cow; let slice = [1u32, 2, 3, 4, 5]; let serialized = to_bytes(&Cow::Borrowed(&slice[..])).unwrap(); let deserialized: Cow<'static, Vec> = from_bytes(&serialized).unwrap(); { let sb: &[u32] = &deserialized; assert_eq!(slice, sb); } let vecx: Vec = deserialized.into_owned(); assert_eq!(slice, vecx[..]); } #[test] fn path_buf() { use std::path::{Path, PathBuf}; let path = Path::new("foo").to_path_buf(); let encoded = to_bytes(&path).unwrap(); let decoded: PathBuf = from_bytes(&encoded).unwrap(); assert!(path.to_str() == decoded.to_str()); } #[derive(Arbitrary, Debug, Deserialize, Serialize, PartialEq)] struct Addr([u8; 32]); #[derive(Arbitrary, Debug, Deserialize, Serialize, PartialEq)] struct Bar { a: u64, b: Vec, c: Addr, d: u32, } #[derive(Arbitrary, Debug, Deserialize, Serialize, PartialEq)] struct Foo { a: u64, b: Vec, c: Bar, d: bool, e: BTreeMap, Vec>, } #[test] fn serde_known_vector() { let b = Bar { a: 100, b: vec![0, 1, 2, 3, 4, 5, 6, 7, 8], c: Addr([5u8; 32]), d: 99, }; let mut map = BTreeMap::new(); map.insert(vec![0, 56, 21], vec![22, 10, 5]); map.insert(vec![1], vec![22, 21, 67]); map.insert(vec![20, 21, 89, 105], vec![201, 23, 90]); let f = Foo { a: u64::max_value(), b: vec![100, 99, 88, 77, 66, 55], c: b, d: true, e: map, }; let bytes = to_bytes(&f).unwrap(); let test_vector = vec![ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x06, 0x64, 0x63, 0x58, 0x4d, 0x42, 0x37, 0x64, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x09, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x63, 0x00, 0x00, 0x00, 0x01, 0x03, 0x01, 0x01, 0x03, 0x16, 0x15, 0x43, 0x03, 0x00, 0x38, 0x15, 0x03, 0x16, 0x0a, 0x05, 0x04, 0x14, 0x15, 0x59, 0x69, 0x03, 0xc9, 0x17, 0x5a, ]; // make sure we serialize into exact same bytes as before assert_eq!(test_vector, bytes); // make sure we can deserialize the test vector into expected struct let deserialized_foo: Foo = from_bytes(&test_vector).unwrap(); assert_eq!(f, deserialized_foo); } #[derive(Debug, Deserialize, Serialize, PartialEq, Eq, Clone)] struct List { next: Option<(usize, Box)>, } impl List { fn empty() -> Self { Self { next: None } } fn cons(value: usize, tail: List) -> Self { Self { next: Some((value, Box::new(tail))), } } fn integers(len: usize) -> Self { if len == 0 { Self::empty() } else { Self::cons(len - 1, Self::integers(len - 1)) } } } #[test] fn test_recursion_limit() { let l1 = List::integers(4); let b1 = to_bytes(&l1).unwrap(); assert_eq!( b1, vec![ 1, 3, 0, 0, 0, 0, 0, 0, 0, 1, 2, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ] ); assert_eq!(from_bytes::(&b1).unwrap(), l1); let l2 = List::integers(MAX_CONTAINER_DEPTH - 1); let b2 = to_bytes(&l2).unwrap(); assert_eq!(from_bytes::(&b2).unwrap(), l2); let l3 = List::integers(MAX_CONTAINER_DEPTH); assert_eq!( to_bytes(&l3), Err(Error::ExceededContainerDepthLimit("List")) ); let mut b3 = vec![1, 243, 1, 0, 0, 0, 0, 0, 0]; b3.extend(b2); assert_eq!( from_bytes::(&b3), Err(Error::ExceededContainerDepthLimit("List")) ); let b2_pair = to_bytes(&(&l2, &l2)).unwrap(); assert_eq!( from_bytes::<(List, List)>(&b2_pair).unwrap(), (l2.clone(), l2.clone()) ); assert_eq!( to_bytes(&(&l2, &l3)), Err(Error::ExceededContainerDepthLimit("List")) ); assert_eq!( to_bytes(&(&l3, &l2)), Err(Error::ExceededContainerDepthLimit("List")) ); assert_eq!( to_bytes(&(&l3, &l3)), Err(Error::ExceededContainerDepthLimit("List")) ); }