Base64.encodeBase64() throws NegativeArraySizeException on large files

• Key: CODEC-61
• Type: Bug
• Status: Closed
• Resolution: Fixed
• Priority: Major
• Assignee: Unassigned
• Reporter: Igor Slepchin
• Affects Version/s: 1.3
• Fix Version/s: 1.4

Environment

java -version
java version "1.5.0_03"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_03-b07)
Java HotSpot(TM) Server VM (build 1.5.0_03-b07, mixed mode)

uname -a
Linux HP-FC7-IGOR 2.6.23.8-34.fc7 #1 SMP Thu Nov 22 20:39:56 EST 2007 x86_64 x86_64 x86_64 GNU/Linux

Description

The NegativeArraySizeException exception is thrown by Base64.EncodeBase64() for arrays larger than 268435455 bytes (2^31/8-1).

public static byte[] encodeBase64(byte[] binaryData, boolean isChunked) starts with the following three lines:

int lengthDataBits = binaryData.length * EIGHTBIT;
int fewerThan24bits = lengthDataBits % TWENTYFOURBITGROUP;
int numberTriplets = lengthDataBits / TWENTYFOURBITGROUP;

The first of the lines will cause an integer overflow in lengthDataBits for lengths larger than 2^31/8-1, making it a negative number. The fix is trivial (but not tested on the running code, I just ran through a few numbers to validate that it computes the same results as the original code):

int lengthData = binaryData.length;
int fewerThan24bits = lengthData % (TWENTYFOURBITGROUP / EIGHTBIT) * EIGHTBIT;
int numberTriplets = lengthData / (TWENTYFOURBITGROUP / EIGHTBIT);

This way the encoder will be able to process files of up to 2^31-1 bytes in length, which is much better than ~250MB.

The issue was found in commons 1.3; the source code above was taken from SVN trunk so I assume it's still present in 1.4: http://svn.apache.org/repos/asf/commons/proper/codec/trunk/src/java/org/apache/commons/codec/binary/Base64.java