infozip also has problems with japanese characters. I ended up solving the problem with the python zip module. considerably more awkward to use but it offers a

infozip also has problems with japanese characters. I ended up solving the problem with the python zip module. considerably more awkward to use but it offers a lot of control over the extraction process. I did not look it very closely so I don't know what exactly infozip was getting incorrect in my case. but I did find this interesting bug report from 2012. Apparently a lot of encoders are sloppy about the spec and will leave header fields zeroed rather than set them. and if infozip reads a header that says the zipfile was created by dos(a zero) it believes it and extracts it using a dos compatible encoding. https://sourceforge.net/p/infozip/support-requests/10/ https://sourceforge.net/p/infozip/support-requests/10/