Fix up a variety of unicode/bytes issues with SVN.
Review Request #9627 — Created Feb. 12, 2018 and submitted — Latest diff uploaded
The SVN client was handling a variety of things using the wrong string
types, most notably filenames. In Python 3, filenames are expected to be
treated asstr
internally, and then serialized/deserialized using the
filesystem encoding (theoretically this is also the case in Python 2,
but the implicit casting makes it easy to be sloppy). This change fixes
those up so that we're at least sucessfully running the tests.This also incidentally fixes bug 4547, which involves using
Popen.communicate()
instead ofPopen.wait()
.
Ran unit tests on Python 2.7 and 3.6.