Block a user
_applet-grid-list.html (newly extracted) so item rendering + empty-state row + scroll-buffer all live in one place — _applet-list-shell.html (the dedicated billbuds/billposts page shell) now internally includes the same grid-list partial for its inner <ul>, so the dedicated-page and in-grid lists share the same skeleton; new per-applet item partials _my_buds_applet_item.html (mirrors _my_buds_item.html w. data-bud-id + display_name), _my_notes_item.html (links to billboard:my_notes; uses display_name), _my_posts_applet_item.html (Post link + title), _my_scrolls_item.html (Room link to billboard:scroll), _my_games_item.html (Room link to epic:gatekeeper); view-side _billboard_context gains _recent_buds(user) — sorts the User.buds auto-through table by -id so newest-added-first w.o. an explicit through model w. timestamps (manage [r.to_user for r in rows]) — + _recent_notes(user) (user.notes.order_by('-earned_at')[:limit]); same two helpers threaded into new_post's GET-with-form-errors branch (line 270-274) so the rerender keeps the new applet content visible; 7 ITs added to BillboardViewTest covering recent_buds ordering / cap / empty + recent_notes ordering / cap / cross-user isolation / empty; SCSS — .applet-list / .applet-list-entry / .applet-list-buffer lifted from .applet-list-page .applet-scroll scope to top level so they apply in both surfaces; in-grid applets get display: flex; flex-direction: column; .applet-list { flex: 1 } so the list scrolls within the applet box; #id_applet_my_games ul-centring + .scroll-list + #id_applet_notes h2 { writing-mode: vertical-rl ... } overrides removed (centring was an empty-state-only behaviour, scroll-list + vertical-rl redundant w. the new shared rule + the %applet-box > h2 rule); My Games items now left-aligned by default; empty-state row recovers the centred-italic-dim treatment via .applet-list-entry--empty { flex: 1; display: flex; align-items: center; justify-content: center; opacity: 0.6; font-style: italic } + .applet-list:has(> .applet-list-entry--empty) { display: flex; flex-direction: column } — so "No buds yet" / "No notes yet" / "No games yet" / "No scrolls yet" / "No posts yet" all centre in their applet aperture, reverting to the left-aligned stack the moment a real item lands; Most Recent Scroll's outer empty <p><small>No recent activity.</small></p> adopts the same .applet-list-entry .applet-list-entry--empty classes (section is already flex-column from existing rule) so it picks up the unified centred-italic-dim treatment
_post_gear.html partial gates DEL/BYE on viewer_is_owner (set by view_post since the buds sprint) + post.kind, then includes the shared apps/applets/_partials/_gear.html btn; styling rides the existing applets.scss page-level pattern — .post-page joins .billboard-page / .room-page / .dashboard-page / .wallet-page / .gameboard-page / .billscroll-page in the > .gear-btn { position: fixed; bottom: 4.2rem; right: 0.5rem } rule (and the landscape footer-sidebar centred variant), #id_post_menu joins the %applet-menu extension list + the page-level fixed-menu rule (bottom: 6.6rem; right: 1rem); 5 FTs in test_bill_post_gear.py (owner DEL flow, invitee BYE flow, 3 menu-shape assertions for owner/invitee/admin) + 11 ITs across DeletePostViewTest + AbandonPostViewTest (302 redirect target, side effect, GET-is-no-op, non-owner / non-invitee / NOTE_UNLOCK protection) — TDD
@<handle>'s Posts → Posts by Me — symmetrises w. the right pane's Posts by Others (both panes now read as point-of-view labels rather than one identity-stamped + one generic), and drops the @<handle> reveal from a screen the owner alone visits (handle is already in the navbar avatar tooltip); handle = owner.username or owner.email derivation goes w. it since nothing else in the view referenced it
DATABASE_URL to the postgres service), so running them concurrently collided on src/test_db.sqlite3: the second-to-start container hit a half-created DB and the runner's Type 'yes' to delete the existing test database prompt EOFed under non-interactive CI stdin (pipeline run #296); fix flips test-FTs-room.depends_on from test-two-browser-FTs → test-FTs-non-room so the heavy room cluster strictly follows the non-room bucket; section comment rewritten from "Parallel FT split" → "FT split (sequential for now)" w. the run #296 EOFError documented + two re-parallelisation paths spelled out for later (per-step distinct sqlite paths via DATABASE_URL=sqlite:////tmp/test_db_<bucket>.sqlite3 OR per-step distinct postgres DBs); the two stale "parallel" mentions (collectstatic note in test-two-browser-FTs + inline comment in test-FTs-room) also updated; screendumps + build-and-push depends_on unchanged — Woodpecker resolves the transitive ordering fine
pip install -r requirements.dev.txt step would not satisfy these prod runtime deps if the image lagged); bump requests 2.31.0 → 2.32.5 in requirements.txt to align w. the dev pin (eliminates the version-drift wedge where CI's dev-only install would silently downgrade-or-upgrade vs prod's pin)
test_bud_btn.py references in the test_core_bud_btn.py docstring + test_admin_post_readonly.py comment to point at the new filename; user-driven renames (22 files): test_applet_my_notes/posts → test_bill_my_*, test_applet_new_post[_line_validation] → test_bill_new_post[_line_validation], test_applet_my_sky → test_dash_my_sky, test_applet_palette → test_dash_palette, test_wallet → test_dash_wallet, test_login → test_core_login, test_navbar → test_core_navbar, test_sharing → test_core_sharing, test_layout_and_styling → test_core_styling, test_my_buds → test_bill_my_buds, test_bud_btn → test_core_bud_btn, test_deck_contribution → test_game_room_deck_contrib, test_game_invite → test_game_room_invite, test_room_gatekeeper → test_game_room_gatekeeper, test_room_role_select → test_game_room_select_role, test_room_sea_select → test_game_room_select_sea, test_room_sig_select → test_game_room_select_sig, test_room_sky_select → test_game_room_select_sky, test_room_tray → test_game_room_tray, test_component_tray_tooltip → test_game_room_tray_tooltip; the post_page.py / room_page.py helper modules from the May-12 sprint absorbed the cross-file FT imports that would otherwise have cascade-broken on these renames
python:3.13-slim to gitea.earthmanrpg.me/discoman/python-tdd-ci:latest (same image as the two FT steps) so all 3 steps inherit the requirements.dev.txt deps Dockerfile.ci already pre-installs; all 3 steps switch pip install -r requirements.txt → pip install -r requirements.dev.txt so the install collapses to ~5-10s of "already satisfied" verification per step (vs ~30-60s of unpinned PyPI resolver+download against requirements.txt); ~3-5 min saved per pipeline run; drift safety net preserved — pip still installs deltas if requirements.dev.txt is ahead of the image, so a stale image doesn't break CI, it just runs slower until the image is rebuilt + pushed
_fill_room_via_orm (was in test_room_role_select, imported by test_room_tray + test_deck_contribution + test_game_invite + test_room_gatekeeper + test_room_sig_select), _assign_all_roles (was in test_room_sig_select, imported by test_room_tray + test_deck_contribution), and _equip_earthman_deck (duplicated verbatim in test_room_role_select + test_component_tray_tooltip — the test_component_tray_tooltip copy of _fill_room_via_orm was also a near-duplicate, missing only the gamers-return both call sites discarded anyway); SIG_SEAT_ORDER constant moves along since _assign_all_roles depends on it; mirrors the existing post_page.py pattern; underscored helper names kept so the "test infrastructure, not API" signal survives the relocation; dropped now-unused per-file imports (DeckVariant from test_room_role_select; Note/DeckVariant/TableSeat/TarotCard from test_room_sig_select; GateSlot from test_component_tray_tooltip); regression gate: GatekeeperTest (8 FTs) green via the new helper home; smoke-imports green across all 8 touched modules
PIP_CACHE_DIR: .pip-cache on all 3 test steps (test-UTs-n-ITs / test-two-browser-FTs / test-FTs) — each step currently re-downloads the entire requirements.txt wheel set independently; Woodpecker mounts the workspace across steps in one run, so the wheels populated by step 1 are reused by steps 2-3, saving ~1-2 min of pip-install wall-clock per run (only download time — install still happens fresh per container); pyswiss.yaml untouched (separate pipeline, separate workspace, separate requirements); cache is per-run only — cross-run persistence would need a Woodpecker volume plugin
body.page-billboard / page-dashboard / page-gameboard / page-sky / page-wallet etc., with 5 near-identical html:has(body.page-X) { overflow: hidden } + body.page-X { … } blocks duplicating the same rules; any page that forgot to set page_class in its view context (e.g. epic.tarot_deck — never set) rendered without the aperture, letting applet borders + titles clip past the fixed navbar/footer sidebars at narrower viewports; foundation now universal, page-specific overrides stay scoped — gameboard keeps .container { overflow: clip } (Firefox seat-tooltip scroll-anchoring quirk) + billboard/dashboard/gameboard keep .row { margin-bottom: -1rem } (h2-row tightening); page_class context vars + body class hooks preserved (FTs at test_bud_btn.py:370 / :379 still assert on them); regression gate: 60 layout-sensitive FTs (billboard, my_buds, bud_btn, applet_my_posts, dashboard, wallet, gameboard, layout_and_styling, jasmine) + 43 room FTs (gatekeeper_bud_btn, room_gatekeeper, room_sky_select, sharing) all green
.bud-entry--empty but _applet-list-shell renders it w. the family-agnostic .applet-list-entry--empty class; the selector never matched, so first-bud appends landed alongside the empty row instead of replacing it; existing test_add_bud_via_bud_btn_appends_entry extended to also assert the empty row dismisses post-add (caught the bug — TDD)
{already_present, recipient_display, recipient_user_id}; bud-btn.js branches on already_present → calls new Brief.showDuplicateBanner({display_name, target_selector}) instead of the normal onSuccess append; banner title reads @<username> is already present, NVM dismisses, FYI dismisses AND eases in the .bud-duplicate-flash class (color: var(--terUser); text-shadow: 0 0 .5em var(--ninUser); transition: 600ms) onto the existing element (.bud-entry .bud-name / .post-recipient[data-user-id=…] / .gate-slot.filled[data-user-id=…]); gatekeeper "already present" = recipient is either GateSlot.FILLED + gamer OR has TableSeat OR has a pending RoomInvite (highlight target only set when seated — pending invites have no visible slot); .post-recipient chips + .gate-slot.filled cells gain data-user-id so the FYI selector can find them; my_buds.html now loads note.js via the {% block scripts %} pattern (Brief module is required by the duplicate banner path); bonus: latent test_jasmine.py bug fixed — "0 failures" in result.text matched "10 failures" / "20 failures" / etc, silently passing up to 99 failed specs; replaced w. re.search(r"(?<!\d)0 failures\b", …) (caught my new red specs, would've caught any prior Jasmine regression); 18 new ITs + 10 new Jasmine specs + 3 new FTs (one per panel) — TDD
bindBudBtn({submitUrl, autocompleteUrl?, onSuccess(data)}) so each panel keeps only its markup + an onSuccess callback (post-share: _appendLine + _appendRecipientChip + Brief.showBanner; gatekeeper: Brief.showBanner only; my-buds: _appendBudEntry); autocomplete binding folded into bud-btn.js (drives bindBudAutocomplete when autocompleteUrl is passed) so callers stop repeating the bindBudAutocomplete(...) boilerplate; dead data-share-url/data-invite-url/data-add-url attrs dropped since bindBudBtn takes the URL directly; behavior-preserving — all 21 existing FTs (test_my_buds + test_bud_btn) + 49 ITs (test_buds + test_share_post + test_brief) green
id_bud_suggestions click-outside guard + unused {% load static %} dropped — TDD
date:'g:i A' for the existing lyric_extras.relative_ts filter so same-day Lines render as 11:46 p.m., then collapse to weekday (Sat) past 24h, 07 Mar past a week, 07 Mar 2025 past a year — DRYs onto the single filter scroll already uses (no duplicated bucketing logic); same-day rendering shifts from 11:46 PM to 11:46 p.m. as a side effect of reusing the lowercase g:i a format relative_ts already standardised on — TDD
right: 2.5rem ≥1800px override on kit-btn that left it outboard of gear-btn after the rem-fluid sidebar refactor