CI #346 test-FTs-room had 4 consistent fails (failed on both the first run
AND the retry, so real, not flakes). All 4 are test-side — the shipped
features are correct; the FTs lagged behind deliberate UX changes + a
race they never needed to depend on.
test_game_my_sea.py
- test_nvm_navigates_back_to_gameboard → renamed test_nvm_navigates_back_to_
my_sea_hex; asserts /gameboard/my-sea/$ now. NVM on the gatekeeper navigates
to the table hex, not out to /gameboard/ (changed 5cade51: gatekeeper +
picker NVM → hex; only landing + sign-gate eject to /gameboard/). Sibling
test_gear_btn_opens_menu_with_nvm_only still passes (only checks the onclick
contains /gameboard/).
- test_default_spread_is_situation_action_outcome → add _open_spread_modal(self).
The spread combobox moved into #id_sea_spread_modal (burger Sea sub-btn
sprint); .sea-select-current .text returned '' while the modal was hidden.
Mirrors the already-updated sibling test_picking_spread_swaps_*.
test_trinket_carte_blanche.py (the recurring #344/#345/#346 carte fail)
- Sign-gate Brief: replace the hard wait_for_slow(find .my-sea-sign-gate-brief)
+ NVM-click with dismiss_brief_if_present(). The Brief fires via
Brief.showBanner on DOM-ready; its appearance is a DOM-ready-vs-note.js-load
race, so under CI contention it sometimes never lands in-window and ANY hard
wait throws NoSuchElement. a39053d misdiagnosed this as a timeout (→
wait_for_slow); it is not. The test never asserts the Brief — it only clears
it to unblock a later click. dismiss_brief_if_present removes it if present +
no-ops if absent: robust to the race.
- Kit-bag token select: JS-click the #id_kit_bag_dialog CARTE token. In
landscape (CI default viewport) the kit-bag dialog is a vertical bar that
slides in via a max-width transition (burger landscape refactor), so a
Selenium .click races the animation + can't scroll the token into the
overflow container ("could not be scrolled into view"). execute_script fires
the bound handler directly. Applied in both carte tests
(open_kit_and_select_carte + the in-use attribution flow); token-rails stays
a normal click (it lives on the gate page, not in the dialog).
Verification: all 4 methods green locally (landscape viewport) —
test_carte_in_use_game_kit_shows_room_attribution 10.8s; the multi-slot carte
+ both my-sea methods 35.5s.
Code architected by Disco DeDisco <discodedisco@outlook.com>
Git commit message Co-Authored-By:
Claude Opus 4.7 <noreply@anthropic.com>
1c799d35ca