Sprint 4a-cont (400762c) + 4b (cd0add1) introduced `serialized_rollback = True` on `MySignPickerTest`, `MySignBackupDeckTest`, and `MySeaSignGateTest` to keep migration-seeded `DeckVariant` + `TarotCard` rows alive across TransactionTestCase flushes. Locally each file ran clean, but in the full pipeline #313 `test-FTs-non-room` stage all three classes errored in setUpClass:
django.db.utils.IntegrityError: UNIQUE constraint failed:
django_content_type.app_label, django_content_type.model
Mechanism: every prior `TransactionTestCase`-derived class in the bucket flushed without `inhibit_post_migrate`, so Django's post-migrate signal recreated `django_content_type` rows. When these three classes hit `_fixture_setup` & tried to deserialize the saved DB snapshot, the inserts collided on `(app_label, model)`. No serialized rollback ⇒ no collision; mixing in the same DB run is the trap.
Fix: drop `serialized_rollback = True` from all 3 classes & inline-reseed via `get_or_create` per the canonical pattern already used by [test_admin_tarot.py](src/functional_tests/test_admin_tarot.py) & [room_page.py](src/functional_tests/room_page.py) (see [[feedback_transactiontestcase_flush]]) — new `_seed_earthman_sig_pile()` module-level helper in each file restores `DeckVariant(slug='earthman')` + the 16 MIDDLE court cards (Maid/Jack/Queen/King × BRANDS/CROWNS/BLADES/GRAILS) that `personal_sig_cards(user)` returns.
Adjacent bug uncovered once the snapshot reload was gone: [test_game_my_sea.py](src/functional_tests/test_game_my_sea.py) `_seed_gameboard_applets` had been seeding `my-palette` w. `context='gameboard'` — but `my-palette` is not a gameboard applet (it's a dashboard slug; the real palette applet is `palette` per migration 0003). The applets template iterates every applet & includes `apps/<context>/_partials/_applet-<slug>.html`, so seeding the bogus row made /gameboard/ try to load a partial that doesn't exist → TemplateDoesNotExist 500 → `#id_applet_my_sea` never rendered. `serialized_rollback = True` had been masking this because the snapshot restored the migration-correct applet rows (which never had my-palette as a gameboard entry to begin with). Swapped `my-palette` for the real `new-game` gameboard applet & corrected grid-cols/rows on `game-kit` to match the 0003 seed (4×3, not 4×6).
Tests: full sweep across all three classes runs green locally (17/17 in 142s).
Code architected by Disco DeDisco <discodedisco@outlook.com>
Git commit message Co-Authored-By:
Claude Sonnet 4.6 <noreply@anthropic.com>
/gameboard/my-sea/ standalone page + /gameboard/ My Sea applet gated behind User.significator. When no sig is saved, render a Look!-formatted Brief-style line — "Look!—pick your sign before drawing the Sea." — w. BACK (.btn-cancel → /gameboard/) + FYI (.btn-info → /billboard/my-sign/) action buttons in `--terUser` ink ; gate is inline content (not portaled like .note-banner) — it IS the page content until a sig is picked, not a transient nudge ; applet partial mirrors the gate via `{% if not request.user.significator_id %}` w. a `.my-sea-sign-gate--applet` denser variant (just FYI, no BACK — user's already on the gameboard); .my-sea-sign-gate__line shrinks from 1.1rem → 0.85rem + padding from 1.5rem → 0.5rem ; my_sea view passes `user_has_sig = request.user.significator_id is not None` so the standalone template branches at server side (avoids a request.user template-context-processor dependency in the standalone page) ; .woodpecker/main.yaml routes test_game_my_sea.py to the test-FTs-non-room stage by default (FT doesn't yet touch the table hex — Sprint 5+ will bring the hex into my_sea via the same DRY .room-shell stack as my_sign, at which point the file gets moved to test-FTs-room) ; TDD trail — 6 FTs in test_game_my_sea.py covering standalone gate copy + FYI/BACK href targets (T1-T3), with-sig skips gate + renders draw shell (T4), applet mirrors gate w. FYI (T5), applet w. sig falls back to .my-sea-empty (T6); all written red against the un-implemented gate before view/template/SCSS landed ; KNOWN: visual verification on existing admin user (@disco) blocked by lack of a clear-sign affordance (he has a sig saved from Sprint 4a testing); adjacent feature spec'd in [[sprint_my_sea_sign_gate_may19]] memory — likely lands as a CLEAR btn in the picker's saved-sig stage state, deferred to next session
Code architected by Disco DeDisco <discodedisco@outlook.com>
Git commit message Co-Authored-By:
Claude Sonnet 4.6 <noreply@anthropic.com>
