Fix docs_from_attrs truncating mid-UTF-8 codepoint

soroban-sdk-macros/src/doc.rs

@@ -20,9 +20,33 @@
             _ => None,
         })
         .map(|s| s.trim().to_string())
-        .join("\n")
-        .as_bytes()
-        .to_vec();
-    docs.truncate(DOCS_MAX_LEN as usize);
-    docs.try_into().unwrap()
+        .join("\n");
+    // Truncate on a char boundary to avoid splitting multi-byte UTF-8 codepoints.
+    let max = DOCS_MAX_LEN as usize;
+    let safe_len = docs.floor_char_boundary(max);
+    docs.truncate(safe_len);
+    docs.into_bytes().try_into().unwrap()
+}
+
+#[cfg(test)]
+mod test {
+    use super::*;
+    use syn::parse_quote;
+
+    #[test]
+    fn test_truncation_does_not_split_multibyte_utf8() {
+        // (DOCS_MAX_LEN - 1) ASCII bytes followed by 'é' (2 bytes: 0xC3
+        // 0xA9) = DOCS_MAX_LEN + 1 bytes. Truncation at DOCS_MAX_LEN keeps
+        // the 0xC3 but drops the 0xA9, producing invalid UTF-8.
+        let padding = "a".repeat(DOCS_MAX_LEN as usize - 1);
+        let doc_value = format!("{padding}é");
+        let attr: Attribute = parse_quote!(#[doc = #doc_value]);
+        let result = docs_from_attrs(&[attr]);
+        let bytes: Vec<u8> = result.into();
+        assert!(
+            std::str::from_utf8(&bytes).is_ok(),
+            "truncation produced invalid UTF-8: trailing bytes {:?}",
+            &bytes[bytes.len().saturating_sub(4)..]
+        );
+    }
 }