diff options
| author | Christophe Besson <cbesson@gmail.com> | 2026-09-19 17:26:07 +0200 |
|---|---|---|
| committer | Christophe Besson <cbesson@gmail.com> | 2026-09-19 17:26:07 +0200 |
| commit | 95dd3dc13aecec85c2fd72410cd4d1f4ed582dfa (patch) | |
| tree | 05ccb0fe80503acebdfa78830c1b2184964de642 /packages/meshbay-node/tests/test_musicbrainz_enabled_policy.py | |
| parent | 9e7b75bb0f6f6649fb00f2dc97059e90b7d52875 (diff) | |
| download | meshbay-95dd3dc13aecec85c2fd72410cd4d1f4ed582dfa.tar.gz | |
refactor(node): delete the revocation subscriber H4 replaced
`RevocationSubscriber` kept its revocations in two in-memory sets. That is the
defect **H4** names: revocations that live only in memory mean a node restart
silently un-revokes everyone, and group revocations were dropped entirely. The
fix was the persisted `Denylist`, fed from `hub_client.maintain_ws` through
`daemon.py`'s `on_revocation`, which verifies the token against the hub key and
calls `deny_user` / `deny_group`.
So this is not merely unused code. It is the version the register records as
wrong, left beside the one that replaced it — and its docstring still opens with
"Usage in daemon:" and three lines of instructions for wiring it up, which is the
worst shape dead code takes: it reads as the way to do the thing. **L7** is about
exactly this, and the index cache removed two days ago was the same shape.
Certainty before deleting, since that was the condition: no import of it anywhere,
static or dynamic; no entry point; nothing in packaging, the systemd units, the
man page, QE or the documentation; and no test covers it. Every module in the
three packages still imports, ruff passes, and the suite is 2893 — unchanged.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Diffstat (limited to 'packages/meshbay-node/tests/test_musicbrainz_enabled_policy.py')
0 files changed, 0 insertions, 0 deletions