Exception when decoding Jackson-encoded `Base64` binary value in YAML

[tlrx]

Using jackson-dataformats-text-2.9.6 binary values that exceed a given length cannot be decoded back and the following exception is thrown:

com.fasterxml.jackson.core.JsonParseException: Illegal character '\' (code 0x5c) in base64 content

This is related to #62 which forces the Base64Variant to be MIME. When the encoded value exceeds a given size, line seperators are inserted into the encoded value but it then fails to be decoded.

I created the following small test and added it to the SimpleGenerationTest in yaml:

public void testCase() throws Exception {
        final int length = 50;
        final byte[] data = new byte[length];
        Arrays.fill(data, (byte) 1);

        final YAMLFactory yamlFactory = new YAMLFactory();

        final ByteArrayOutputStream os = new ByteArrayOutputStream();
        try (JsonGenerator generator = yamlFactory.createGenerator(os, JsonEncoding.UTF8)) {
            generator.writeStartObject();
            generator.writeBinaryField("data", data);
            generator.writeEndObject();
        }

        try (JsonParser parser = yamlFactory.createParser(os.toByteArray())) {
            assertEquals(JsonToken.START_OBJECT, parser.nextToken());
            assertEquals(JsonToken.FIELD_NAME, parser.nextToken());
            assertEquals("data", parser.currentName());
            assertEquals(JsonToken.VALUE_EMBEDDED_OBJECT, parser.nextToken());
            assertArrayEquals(data, parser.getBinaryValue());
            assertEquals(JsonToken.END_OBJECT, parser.nextToken());
            assertNull(parser.nextToken());
        }
    }

This test passed correctly when length is small (50) but with larger size (like 100) it fails.

[cowtowncoder]

Ok. Thank you for reporting this. Sounds like a bug indeed, related to splitting of base64 encoded lines to conform to 76 character limitation (for some variants).

Not sure what would be the best way to fix, and which part (jackson yaml codec or snakeyaml) is responsible for either unnecessary escaping, or missing decoding.

[vboulaye]

I have the same problem encoding/decoding binary arrays longer than 76 characters after upgrading from 2.9.2 to 2.9.6.

I submitted a PR to try to fix it by changing the way the encoding is done in Base64Variant