Python 3 has utf-8 as a default encoding, so it is worth to add a comment why it is needed here. I assume that it is because that it is not working in case when there is C locale enabled? In such case it would turn to encoding='locale' thus using ascii instead of utf-8.

:thumbsup:

Also, some testing configs would be valuable here.

Dropping testing-ready, currently merge conflict with another testing-ready PR.

Before we go about making these changes, do we have a replicator for #4191 ?

A casual attempt to replicate the complaint ("comment in the config files contains an unicode character"), does not seem to hit an error with the original code.

I've replicated it with such comment in web.conf:

# comment ěščřčž
[web]
SiteName = koji
KojiHubURL = http://koji-hub/kojihub
KojiFilesURL = http://localhost:8080/kojifiles
...

I've replicated it with such comment in web.conf:

I'm still unable to do so. What platform? What locale?

Can you replicate it with a plain call to config.read?

I've tried to make a unit test that manifests this and so far I have failed. The situation is complicated by the various places that we adjust locale in the unit tests without resetting it (and the the difficulty in sanely doing such a reset).

Ok, finally got it. almalinuxorg/8-base container. This hits it

import koji
import locale
locale.setlocale(locale.LC_ALL, 'C')
cfg = koji.read_config_files('test.cfg')

I thought I'd tried this permutation before. I must have done something odd.

additional unit test here -- https://pagure.io/fork/mikem/koji/commits/pr4214updates
(fails on rhel8 if I revert the fix)

Commit 0a7e76d9 fixes this pull-request