array length changes in openjdk

Question

I've recently seen strange behaviour in my application that I can't really explain. They occur very unfrequent so I can't reproduce them. While the place where these problems occur changes, the common part seems to be that an array changes size after it has been created (I know, this is not possible, hence: strange behaviour).

A couple of examples to make my point clear:

java.lang.StringIndexOutOfBoundsException: String index out of range: 86
    at java.lang.String.checkBounds(String.java:409)
    at java.lang.String.<init>(String.java:577)
    at com.acunia.fleet.diagnostics.providers.tacho.VDOKLineInputParser.getRealDriverID(Unknown Source)

The code that would cause this:

public String getRealDriverID(byte[] buffer) {
  if (buffer.length > 86 && isDriverCardInserted(buffer)) {
    return new String(buffer, 70, 16);
  }   
  return null;
}

So we first check that the buffer is big enough (more than 86 bytes) before trying to create a String from them.

A second example:

java.lang.ArrayIndexOutOfBoundsException: -1
    at java.lang.String.lastIndexOf(String.java:1889)
    at java.lang.String.lastIndexOf(String.java:1835)
    at java.lang.String.lastIndexOf(String.java:1817)
    at com.acunia.service.position.nmea.comm.CommPositionProvider.isValid(Unknown Source)

The line that causes this exception is:

int csi = line.lastIndexOf("*");

I had a look at the openjdk String.java source, but couldn't find a fault unless arrays could suddenly start changing size after they where created.

The only reference online that I could find that might be related was as a openjdk bug: https://bugs.openjdk.java.net/browse/JDK-6817012. This bug was marked a 'not an issue' though, although when reading this I can't really tell if the problem that is mentioned is not recognized as a bug, or if the person who closed the bug doesn't see why this bug would cause problems.

If anyone has ever encountered a similar problem I would really appreciate hearing about it. As it stands the problem is too unstable to attempt to fix it by using other versions of openjdk.

Problem was seen on:

hardware: custom arm platform
java version "1.6.0_31"
OpenJDK Runtime Environment (IcedTea6 1.13.3) (6b31-1.13.3-1~deb7u1)
OpenJDK Zero VM (build 23.25-b01, mixed mode)
Linux 3.2.0 #1 Fri Jun 20 10:25:16 CEST 2014 armv7l GNU/Linux

Answer 1

This is most likely an issue with the bytes and the system charset. The documentation explicitly says:

The behavior of this constructor when the given bytes are not valid in the default charset is unspecified.

You can circumvent this by explicitly providing the correct charset:

new String(buffer, 70, 16, StandardCharsets.UTF_8)

Answer 2

You are creating the new String using byte[] array without specifying the Charset. Thus my guess is that system default charset got changed and you see the behaviour changes. I'd suggest to always specify the charset explicitly. For example:

return new String(buffer, 70, 16, "UTF-8");