Python OpenID error: No matching endpoint found after discovering [uri]

Question

Hello,

I'm getting a "No matching endpoint found after discovering [openid identifier]" error when I invoke complete() method of the consumer.

The funny thing is that out of the four OpenID providers I have tested, this behavior is observed only with LiveJournal. What steps can you suggest to investigate and fix the problem?

store = FileOpenIDStore("/path/to/store")

def login(req, uri):
    req.content_type = "text/html"
    session = Session.Session(req)
    consumer = Consumer(session, store)
    auth = consumer.begin(uri)
    util.redirect(req, auth.redirectURL("http://example.com", "http://example.com/authtest.py?sid=" + session.id()))
    return

def index(req, sid):
    req.content_type = "text/html"
    c = Consumer(Session.Session(req, sid), store)
    args = req.args.split("&")
    arg_dict = {}
    for i in range(0, len(args)):
        x, y = args[i].split("=")
        arg_dict[x] = unquote(y)
    v = c.complete(arg_dict, "http://example.com/authtest.py?" + req.args)
    if v.status == 'failure':
        return v.message
    else:
        return v.status

Answer 1

I don't see any glaring error in your code, but here are some steps to investigate:

• Is there any output from oidutil.log? It logs to stderr by default, but you can override it to log to somewhere else if your web server doesn't let you see stderr.

• Capture all requests/responses. You can use something like TamperData to get the indirect request/response passed through the browser, and feed it to contrib/openid-parse from the python-openid distribution to make it more something readable.

• Does the example consumer in the python-openid source distribution work with your LJ identifier? If so, what are the differences in request/response between the example and your code?

• Does your LJ identifier have any punctuation in it?

• Is LJ the only OpenID version 1.x provider you're testing against? (Probably. Hopefully there aren't too many left.)

• Your argument parsing could use urlparse.parse_qs, but I'm not sure that's really a problem. (And parse_qs returns {key: [list-of-values]}, whereas Consumer.complete expects {key: single-value}, so you have to map one to the other.)