Comment 7 for bug 1363348

Django's multipartparser splits incoming data into chunks to process it. On my system, these chunk sizes are 64kb at largest, but are always a multiple of 4 when decoding base64 data, since base64 encoded data must be split at 4 byte boundaries. django is taking the len() of the string of data and making sure that's a multiple of 4, and passing the string to b64decode to decode.

I believe the problem is the data apiclient uploads includes CRLF characters. Django counts those characters as a part of the string when considering the length of a chunk of base64 data, but the CRLF characters aren't significant in base64, so the resulting pure base64 data, with whitespace stripped, isn't necessarily aligned to a 4 byte boundary.