ebbe.bass/ping-mee-self-service-platform
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor: sweep the platform-and-sign-in epic

Browse Source 
Split load_config into per-section readers, use contextlib.suppress and a single condition where ruff asked, remove stale noqa comments, add the SIM and RUF100 rules, share the sign_in test helper, align AD-6 with the code and record the remaining deferred findings.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N3t4WL7bd9bdHHNjEsxzqC
This commit is contained in:
ebbe.bass
2026-10-18 09:29:59 +02:00
co-authored by Claude Sonnet 5.5
parent 20510a73e5
commit 700295cb72
15 changed files with 229 additions and 164 deletions
Download Patch File Download Diff File Expand all files Collapse all files
_bmad-output/initiative-ping-mee-portal/deferred-work.md
+2 -2
View File
@@ -2,5 +2,5 @@
summary: ARCHITECTURE-SPINE AD-2 says the service account's read rights are limited to a list, but Active Directory gives authenticated users broad read by default, so the statement is not enforced by the test-environment setup.
evidence: The setup script grants only write rights and the Change Password right; read access is the default. Either soften AD-2 ("needs no more than this list") or enforce it with deny entries; decide when the architecture is next updated.
- source_plan: `_bmad-output/initiative-ping-mee-portal/epic-platform-and-sign-in/story-audit-throttle-and-session-foundations-plan.md`
summary: ARCHITECTURE-SPINE AD-6 names `end_sessions(username, except_current=bool)`, but the built function is `end_sessions(user, except_session_key=None)`.
evidence: A function that keeps the current session needs that session's key; a boolean cannot say which one. Reword AD-6 to match when the architecture is next updated; callers use the key form.
summary: Concurrent writers on the throttle counters are not tested.
evidence: The test database is SQLite in memory with a shared cache, which answers "table is locked" to a second writer at once, so a threaded test fails for reasons that say nothing about production (a WAL file database with a 20 s busy timeout and IMMEDIATE transactions). A real test needs a file database; tried and dropped in the refactor sweep.