HeaderUtils.ExtractEntityHeaders() should pad content-md5 headers (if needed) before base64 decoding them

[adennie]

When processing a content-md5 header, ExtractEntityHeaders() does the following

            } else if (header.getName().equalsIgnoreCase(
                    HeaderConstants.HEADER_CONTENT_MD5)) {
                result.setDigest(new org.restlet.data.Digest(
                        org.restlet.data.Digest.ALGORITHM_MD5,
                        org.restlet.engine.util.Base64.decode(header
                                .getValue())));

Since an MD5 hash is 128 bits long, its base64 encoding is 22 bytes if unpadded, or 24 bytes if padded. However, the Base64.decode() method (reasonably) requires its input to be a multiple of 4 bytes:

    if (len % 4 != 0) {
        throw new IllegalArgumentException(
                "Base64.decode() requires input length to be a multiple of 4");
    }

Thus, a fix for this issue would be for ExtractEntityHeaders() to pad a 22-byte content-md5 header value with two base64 padding characters ("==") before passing the value to Base64.decode().

[jlouvel]

That sounds good, could you propose a pull request with this fix?

[thboileau]

Thanks a lot Andy.
I've added your contribution to the 2.1 and master branches.