identifiers: Don't accept colons in user-id localparts, even in compat

crates/ruma-common/CHANGELOG.md

@@ -1,5 +1,10 @@
 # [unreleased]
 
+Bug fixes:
+
+- Don't accept colons in the localpart given to `UserId::parse_with_servername`
+  even with `feature = "compat"`
+
 Improvements:
 
 - Derive `Hash` for `ReceiptType` and `ReceiptThread`

crates/ruma-common/src/identifiers/user_id.rs

@@ -206,6 +206,11 @@ mod tests {
         UserId::parse_arc(user_id).unwrap_err();
     }
 
+    #[test]
+    fn definitely_invalid_user_id() {
+        UserId::parse_with_server_name("a:b", server_name!("example.com")).unwrap_err();
+    }
+
     #[test]
     fn valid_historical_user_id() {
         let user_id =

crates/ruma-identifiers-validation/src/user_id.rs

@@ -23,12 +23,22 @@ pub fn localpart_is_fully_conforming(localpart: &str) -> Result<bool, Error> {
         .bytes()
         .all(|b| matches!(b, b'0'..=b'9' | b'a'..=b'z' | b'-' | b'.' | b'=' | b'_' | b'/'));
 
-    // If it's not fully conforming, check if it contains characters that are also disallowed
+    if !is_fully_conforming {
-    // for historical user IDs. If there are, return an error.
+        // If it's not fully conforming, check if it contains characters that are also disallowed
-    // See https://spec.matrix.org/v1.4/appendices/#historical-user-ids
+        // for historical user IDs. If there are, return an error.
-    #[cfg(not(feature = "compat"))]
+        // See https://spec.matrix.org/v1.4/appendices/#historical-user-ids
-    if !is_fully_conforming && localpart.bytes().any(|b| b < 0x21 || b == b':' || b > 0x7E) {
+        #[cfg(not(feature = "compat"))]
-        return Err(Error::InvalidCharacters);
+        let is_invalid = localpart.bytes().any(|b| b < 0x21 || b == b':' || b > 0x7E);
+
+        // In compat mode, allow anything except `:` to match Synapse. The `:` check is only needed
+        // because this function can be called through `UserId::parse_with_servername`, otherwise
+        // it would be impossible for the input to contain a `:`.
+        #[cfg(feature = "compat")]
+        let is_invalid = localpart.as_bytes().contains(&b':');
+
+        if is_invalid {
+            return Err(Error::InvalidCharacters);
+        }
     }
 
     Ok(is_fully_conforming)