Internal: Base64 decode parsing detects more errors

[erikringsmuth]

I found a bug in the Base64.decode() function. It will decode incorrectly formatted base 64 strings. For example, "user:password" encodes to "dXNlcjpwYXNzd29yZA==". Both "dXNlcjpwYXNzd29yZA==123" and "dXNlcjpwYXNzd29yZA=5" decode to "user:password" although neither of them are valid base 64 strings. You can add or replace characters after the first padding character.

I wrote a quick test to show the bug.

@Test
public void testBase64DecodeWithExtraCharactersAfterPadding() {
    // "user:password" encodes to "dXNlcjpwYXNzd29yZA=="
    String correctlyEncoded = "";
    String extraCharacters = "";
    String replacePaddingCharacter = "";
    try {
        correctlyEncoded = new String(Base64.decode("dXNlcjpwYXNzd29yZA=="));
        extraCharacters = new String(Base64.decode("dXNlcjpwYXNzd29yZA==123"));
        replacePaddingCharacter = new String(Base64.decode("dXNlcjpwYXNzd29yZA=5"));
    } catch (IOException e) {
    }
    assertEquals("user:password", correctlyEncoded);
    assertNotEquals("user:password", extraCharacters); // this assertion fails
    assertNotEquals("user:password", replacePaddingCharacter); // this assertion fails
}

I think this line is the culprit.
https://github.com/elasticsearch/elasticsearch/blob/master/src/main/java/org/elasticsearch/common/Base64.java#L1219

It breaks early when it should really throw an IOException like the code 8 lines down.