I think merging pyjwt 2 broke things

[athornton]

  File "./giftless/view.py", line 72, in post
    response['objects'] = [action(**o) for o in payload['objects']]
  File "./giftless/view.py", line 72, in <listcomp>
    response['objects'] = [action(**o) for o in payload['objects']]
  File "./giftless/transfer/basic_external.py", line 42, in upload
    lifetime=self.VERIFY_LIFETIME)
  File "./giftless/transfer/__init__.py", line 69, in _preauth_headers
    return self._auth_module.preauth_handler.get_authz_header(identity, org, repo, actions, oid, lifetime=lifetime)
  File "./giftless/auth/jwt.py", line 127, in get_authz_header
    token = self._generate_token_for_action(*args, **kwargs)
  File "./giftless/auth/jwt.py", line 154, in _generate_token_for_action
    return self._generate_token(**token_payload).decode('ascii')
AttributeError: 'str' object has no attribute 'decode'

I wasn't getting this before. I haven't looked yet, but I see there's a new major version of pyjwt, and I strongly suspect something here changed a type from bytes[] to str. I'll take a look when I have time.

[athornton]

return jwt.encode(payload, self.private_key, algorithm=self.algorithm, headers=headers) # type: ignore

Looks very suspicious. Betting this went from bytes to str in pyjwt 2.x. We should probably get rid of the type:ignore and
do the casting properly.

[athornton]

Yep:

https://pyjwt.readthedocs.io/en/stable/changelog.html#v2-0-0

jwt.encode(...) return type

Tokens are returned as string instead of a byte string

[athornton]

And typing in pyjwt turns out to be a giant crawling can of worms. cf jpadilla/pyjwt#602

[rufuspollock]

@athornton this should be fixed i guess now we merged #124. Can you confirm and then we can close this.

[athornton]

I can confirm, but not till next week (Jan. 2, 2024), when I'm back at work.

[athornton]

Oh, whoops, should have given the OK to close a while ago. This (indeed a later merge that includes this) is working fine.

[rufuspollock]

FIXED.