Object changes are not logged after upgrade to v4.1.9

[koratfood]

Deployment Type

Self-hosted

Triage priority

N/A

NetBox Version

v4.1.9

Python Version

3.12

Steps to Reproduce

1. [v4.1.8] Create site "Testsite1" with slug "testsite1" and otherwise default values
2. [v4.1.8] Edit site "Testsite1", enter "hello" in the Description field, and save
3. [v4.1.8] Delete site "Testsite1"
4. Upgrade to v4.1.9
5. [v4.1.9] Create site "Testsite2" with slug "testsite2" and otherwise default values
6. [v4.1.9] Edit site "Testsite2", enter "hello" in the Description field, and save
7. [v4.1.9] Delete site "Testsite2"

Expected Behavior

All 6 changes are found in the Changelog.

Observed Behavior

Only changes made in reproduction steps 1-3 (i.e. before we upgraded to v4.1.9) are found in the Changelog.

[sleepinggenius2]

I can confirm the same issue. For some additional context, starting with a v4.1.9 environment, simply a git checkout v4.1.8 and restarting the service without running update.sh is sufficient to restore changelog functionality. I have also confirmed that this applies to changes made through the UI, scripts, and API.

[tobiasge]

I noticed the same. The problem was likely introduced in this PR: #18225
The event_tracking request processor is not put into the registry.

From our installation:

./netbox/manage.py shell
Python 3.12.8 (main, Dec  4 2024, 06:51:15) [GCC 14.2.1 20241116] on linux
Type "help", "copyright", "credits" or "license" for more information.
(InteractiveConsole)
>>> from netbox.registry import registry
>>> registry.keys()
dict_keys(['counter_fields', 'data_backends', 'denormalized_fields', 'event_types', 'model_features', 'models', 'plugins', 'request_processors', 'search', 'tables', 'views', 'widgets'])
>>> registry.get('data_backends')
{'local': <class 'core.data_backends.LocalBackend'>, 'git': <class 'core.data_backends.GitBackend'>, 'amazon-s3': <class 'core.data_backends.S3Backend'>}
>>> registry.get('request_processors')
[]

[layer3lv]

Same problem in docker version

[abhi1693]

Webhooks are also not triggered. The related PR fixes that too.

[jeremystretch]

I tracked down why this regression was not caught by CI for PR #18225. When the change logging tests are run in isolation, they fail:

$ ./manage.py test --keepdb core.tests.test_changelog
Found 13 test(s).
Using existing test database for alias 'default'...
System check identified no issues (0 silenced).
FFFEEEEEEEEF.
...

This is because the event_tracking context manager is never imported, and thus does not get registered (which PR #18262 fixes).

However, when the entire test suite is run (as during CI), event_tracking is coincidentally imported by a test module for event rules, which happens to run before the change logging tests.

Reliably checking for the registration of event_tracking in a test might prove difficult. Will have to give it some more thought.