fix: should validate when parsing into lazyvalue

Cargo.toml

@@ -10,7 +10,7 @@ license       = "Apache-2.0"
 name          = "sonic-rs"
 readme        = "README.md"
 repository    = "https://github.com/cloudwego/sonic-rs"
-version       = "0.4.0-rc1"
+version       = "0.4.0-rc2"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 

fuzz/Cargo.toml

@@ -16,6 +16,7 @@ faststr       = "0.2"
 libfuzzer-sys = "0.4"
 serde         = { version = "1.0", features = ["derive"] }
 serde_json    = { version = "1.0", features = ["float_roundtrip"] }
+simdutf8      = "0.1"
 sonic-rs      = { path = ".." }
 
 [[bin]]

fuzz/src/lib.rs

@@ -111,9 +111,22 @@ pub fn sonic_rs_fuzz_data(data: &[u8]) {
                 }
             }
         }
-        Err(_) => {
-            let _ = from_slice::<Value>(data)
-                .expect_err(&format!("parse invalid json {:?} failed", data));
+        Err(e) => {
+            let _ = from_slice::<Value>(data).expect_err(&format!(
+                "parse invalid json {:?} failed, should return error {e} ",
+                data
+            ));
+
+            // LazyValue should return error if the json is invalid
+            let msg = e.to_string();
+            if (msg.starts_with("expected ") || msg.starts_with("EOF"))
+                && simdutf8::basic::from_utf8(data).is_ok()
+            {
+                let _ = from_slice::<OwnedLazyValue>(data).expect_err(&format!(
+                    "parse invalid json {:?} failed, should return error {msg}",
+                    data
+                ));
+            }
         }
     }
 

src/lazyvalue/owned.rs

@@ -890,4 +890,21 @@ mod test {
         assert_eq!(to_string(&root).unwrap(), to_string(&root2).unwrap());
         assert_eq!(to_string(&root).unwrap(), to_string(&root3).unwrap());
     }
+
+    #[test]
+    fn test_owned_from_invalid() {
+        for json in [
+            r#"", 
+            r#"{"a":"#,
+            r#"{"a":123"#,
+            r#"{"a":}"#,
+            r#"{"a": x}"#,
+            r#"{"a":1.}"#,
+            r#"{"a:1.}"#,
+            r#"{"a" 1}"#,
+            r#"{"a"[1}}"#,
+        ] {
+            crate::from_str::<OwnedLazyValue>(json).expect_err(json);
+        }
+    }
 }

src/parser.rs

@@ -581,28 +581,34 @@ where
     #[inline(always)]
     pub(crate) fn get_owned_lazyvalue(&mut self, strict: bool) -> Result<OwnedLazyValue> {
         let c = self.skip_space();
-        let start = self.read.index() - 1;
-        match c {
-            Some(b'"') => match self.skip_string()? {
-                ParseStatus::None => {
-                    let slice = self.read.slice_unchecked(start, self.read.index());
-                    let raw = unsafe { self.read.slice_ref(slice).as_faststr() };
-                    return Ok(OwnedLazyValue::from_non_esc_str(raw));
+        let start = match c {
+            Some(b'"') => {
+                let start = self.read.index() - 1;
+                match self.skip_string()? {
+                    ParseStatus::None => {
+                        let slice = self.read.slice_unchecked(start, self.read.index());
+                        let raw = unsafe { self.read.slice_ref(slice).as_faststr() };
+                        return Ok(OwnedLazyValue::from_non_esc_str(raw));
+                    }
+                    ParseStatus::HasEscaped => {}
                 }
-                ParseStatus::HasEscaped => {}
-            },
+                start
+            }
             Some(b't') if self.match_literal("rue")? => return Ok(true.into()),
             Some(b'f') if self.match_literal("alse")? => return Ok(false.into()),
             Some(b'n') if self.match_literal("ull")? => return Ok(().into()),
+            None => return perr!(self, EofWhileParsing),
             _ => {
+                let start = self.read.index() - 1;
                 self.read.backward(1);
                 if strict {
                     self.skip_one()?;
                 } else {
                     self.skip_one_unchecked()?;
                 }
+                start
             }
-        }
+        };
         let end = self.read.index();
         let sub = self.read.slice_unchecked(start, end);
         let raw = unsafe { self.read.slice_ref(sub).as_faststr() };

src/serde/de.rs

@@ -373,7 +373,7 @@ impl<'de, R: Reader<'de>> Deserializer<R> {
     where
         V: de::Visitor<'de>,
     {
-        let val = ManuallyDrop::new(self.parser.get_owned_lazyvalue(false)?);
+        let val = ManuallyDrop::new(self.parser.get_owned_lazyvalue(true)?);
         // #Safety
         // the json is validate before parsing json, and we pass the document using visit_bytes
         // here.