As discovered by #4258, in our current implementation of rzip.CreateFromDirectory, the following behaviors are observed:

1. Symbolic-linked files are followed, and the target contents are copied into the archive.
2. The function throws an error "read: some_file_path: is a directory` while processing directories containing symbolic directories.

Here's the failing test for that. Comment out line 52-55 to get it into a partial failure to observe the symbolic-linked files being followed.

A tactical thing we could do in the short-term is to add a skip for symbolic-linked directories. This would at least allow the archive processing to continue, while preserving the existing behavior.

The "right" thing to do is probably to recursively follow symbolic-linked directories, with a depth limit to bail out to avoid infinite recursion (aligned with gnuutils: zip behavior).