Doesn't properly handle non-ASCII UTF-8 characters in GELF input

[mikaelstaldal]

• Version: Logstash 5.3.2
• Operating System: Ubuntu 16.04
• Config File:

input { gelf { } }
output {
  elasticsearch { hosts => ["localhost:9200"] }
  stdout { codec => json_lines }
}

• Sample Data:

{"version":"1.1","host":"udp-zlib","timestamp":1493384075.248,"level":6,"_thread":"main","_logger":"HelloWorld","_additionalField1":"constant value","_additionalField2":"foo bar","_bar":"BAR","_foo":"FOO","short_message":"Hello, world! åäö 1"}

When sending in the above GELF message (compressed with ZLIB, over UDP) to Logstash (using Log4j 2.8.2), the non-ASCII characters in short_message gets garbled. It ends up like this in ElasticSearch:

 {
          "source_host": "127.0.0.1",
          "level": 6,
          "logger": "HelloWorld",
          "foo": "FOO",
          "thread": "main",
          "message": "Hello, world! åäö 1",
          "version": "1.1",
          "bar": "BAR",
          "@timestamp": "2017-04-28T10:28:06.263Z",
          "host": "udp-zlib",
          "@version": "1",
          "additionalField1": "constant value",
          "additionalField2": "foo bar"       
}

It seems that Logstash GELF input doesn't decode the GELF message with UTF-8 as it should according to GELF spec.