Comment 2 Felix Botner 2021-09-01 10:08:13 CEST

Works in my test setup, reverted workaround.

What i always see is

[2021-08-23 12:31:40 +0200] [13715] [DEBUG] Couldn't get username: 'utf-8' codec can't decode byte 0xa7 in position 21: invalid start byte
[2021-08-23 12:31:41 +0200] [13714] [DEBUG] GET /v1/imports/users/1/

in /var/log/univention/ucs-school-import/gunicorn_error.log, but i think that is just an issue with the logging in gunicorn

our http api client uses requests for the communication, and here the password is latin-1

 requests/vi auth.py
    if isinstance(password, str):
        password = password.encode('latin1')

the api server correctly decodes this:

# Header encoding (see RFC5987)
HTTP_HEADER_ENCODING = 'iso-8859-1'
/usr/lib/python3/dist-packages/rest_framework/authentication.py
   auth_parts = base64.b64decode(auth[1]).decode(HTTP_HEADER_ENCODING).partition(':')

but gunicron uses utf-8.

/usr/lib/python3/dist-packages/gunicorn/glogging.py
  auth = base64.b64decode(auth[1].strip().encode('utf-8'))

Which results in this error during the logging. Not sure what to do, ignoring or fix?

Comment 3 Florian Best 2021-09-01 10:17:18 CEST

(In reply to Felix Botner from comment #2)
All right, then this was an interim problem due to Bug #53451.

> Which results in this error during the logging. Not sure what to do,
> ignoring or fix?

Ignoring. it's only logging.

Comment 4 Florian Best 2021-09-02 16:58:20 CEST

OK: changes reverted
OK: test case passes again
except for https://jenkins.knut.univention.de:8181/job/UCSschool-5.0/job/Upgrade%20Singleserver/lastCompletedBuild/Config=s4,TestGroup=import1/testReport/90_ucsschool/302_http-api_class_column_no_school/master206/ - but this is another reason.