Problem with protobuf generation when there is a multiline comments in the msg file

[sebastian-freitag]

ROS2 Galactic.

Simple Example of the Bug.

Consider the file UUID.msg, identical with ros2/unique_identifier_msgs/msg/UUID.msg:

# A universally unique identifier (UUID).
#
#  http://en.wikipedia.org/wiki/Universally_unique_identifier
#  http://tools.ietf.org/html/rfc4122.html

uint8[16] uuid

Consider the file UUID.idl, generated by ROS2:

// generated from rosidl_adapter/resource/msg.idl.em
// with input from unique_identifier_msgs/msg/UUID.msg
// generated code does not contain a copyright notice


module unique_identifier_msgs {
  module msg {
    typedef uint8 uint8__16[16];
    @verbatim (language="comment", text=
      " A universally unique identifier (UUID)." "\n"
      "" "\n"
      "  http://en.wikipedia.org/wiki/Universally_unique_identifier" "\n"
      "  http://tools.ietf.org/html/rfc4122.html")
    struct UUID {
      uint8__16 uuid;
    };
  };
};

Consider the file UUID.proto, this is the file generated by rosidl_typesupport_protobuf/rosidl_adapter_proto/bin/rosidl_adapter_proto:

// generated from rosidl_adapter_proto/resource/idl.proto.em
// with input from unique_identifier_msgs:msg/UUID.idl
// generated code does not contain a copyright notice

syntax = "proto3";

package unique_identifier_msgs.msg.pb;

// A universally unique identifier (UUID).

 http://en.wikipedia.org/wiki/Universally_unique_identifier
 http://tools.ietf.org/html/rfc4122.html
message UUID
{

 bytes uuid = 293563397;
}

This is wrong and protoc will not compile it.
Lines 9-12 should instead be this:

// A universally unique identifier (UUID).
//
//  http://en.wikipedia.org/wiki/Universally_unique_identifier
//  http://tools.ietf.org/html/rfc4122.html

I think the problem is with this expression inside msg.proto.em: comment = "//" + re.sub(r"\\n", "\n//", annotation.value["text"]). However, I do not know why this ever worked, maybe re changed or an issue about different python versions. On my environment, if I change this line to comment = "//" + re.sub(r"\n", "\n//", annotation.value["text"]), then generation works OK.

[r7vme]

@sebastian-freitag-pylot we were also hit by this bug. It also was not clear for me if smth changed in re behaviors or it never worked.