Prior to protobuf 3.12, when parsing a map using TextFormat:

• If there are duplicate map keys, the last key seen is used.
• toString emits the duplicate keys in the original order, so if the ascii representation was parsed again, the same last key will be used, and the original message will be preserved.

In 3.12.0-rc-1 and 3.12.0-rc-2:

• The last key seen is used.
• toString does not retain the original order. In some cases, the last value printed for a key, is not necessarily the value that is being used in a message. As a result, parsing again the output of toString results in a different message.

In other words, m.toString() is producing a text representation that when parsed does not give the original m back.

What version of protobuf and what language are you using?
Version: 3.12.0-rc-2
Language: Java
What operating system (Linux, Windows, ...) and version?
Ubuntu 19.10

What did you do?

np.proto:

syntax = "proto3";

package mypkg;

message Msg {
  map<int32, int32> m = 1;
}

Main.java:

import mypkg.Np.Msg;

import com.google.protobuf.TextFormat;
import com.google.protobuf.util.JsonFormat;

import java.lang.String;

public class Main {
    public static void main(String[] args) throws Exception {
            String input =
            "m: {\n" +
            "  key: 1\n" +
            "  value: 1\n" +
            "}\n" +
            "m: {\n" +
            "  key: -2147483647\n" +
            "  value: 5\n" +
            "}\n" +
            "m: {\n" +
            "  key: 1\n" +
            "  value: -1\n" +
            "}\n";

            Msg p = TextFormat.parse(input, Msg.class);
            int i1 = p.getMMap().get(1);
            Msg p2 = TextFormat.parse(p.toString(), Msg.class);
            int i2 = p2.getMMap().get(1);
            System.out.println(i1);  // prints -1
            System.out.println(i2);  // prints 1, expected -1
    }
}

What did you expect to see
Either a failure to parse, or:

-1
-1

What did you see instead?

1
-1