decode base64: invalid input

That version will not decode (by default) lines with separators, yet the encoder does that by default. (Newer versions don’t have this problem.)

One solution:

base64 -w 0 foo.zip | base64 -d > foo2.zip

Alternate:

base64 foo.zip | base64 -di > foo2.zip

The -i option stands for (from the man page):

-i, --ignore-garbage
       When decoding, ignore non-alphabet characters.
[...]
Decoding require compliant input by default, use --ignore-garbage to
attempt to recover from non-alphabet characters (such as newlines)