Ordered by priority. Updated end of session 13z8 (2026-06-05).

Structural blockers are solved (the dataExpanded chain + the QML-GC use-after-free class). We're in a UI-polish phase. Authoritative open-issues list: status.md → "Open Issues". History: sessions/. UI conventions: reference/ui-patterns.md.

Done — two big screens polished + iterated

Pokémon details editor (13k–13t): responsive (function onX() → onX:), proper layouts, borderless combos w/ hover underlines, square move pills, slider tooltips, popup nickname editor.
Name editors — full keyboard + quick-edit popup (13v–13z8): full rebuild + many live iterations. Simulated group (Outdoor/Tileset/Grid-Tileset, FlatToggle), single-select radio filters, pill grid
- image-only static tile tooltip, ⋮ menus replaced by buttons (dice randomize, Name/Example toggle + >>), example demo localized to each editor, plus the player/rival/ID edit-hang + OT-corruption fix and the expandStr infinite-loop + tileset off-by-one fixes. Conventions: ui-patterns.md.

Now / next

Continued review of the name editors. Live tweaks ongoing. NameDisplay/NameEdit are shared by player/rival/nickname + the keyboard footer — verify all on each rebuild.
Decide the fate of the right-side DetailView in the full keyboard. The pill tooltip now shows the rendered tile; DetailView still shows text info on hover (and the tileset page uses it). Ask whether to keep it, slim it, or drop it for more grid room.
End-to-end persistence pass (edit → save → reopen) across player, rival, money, coins, badges, pokédex, items, pokémon. several have been spot-confirmed; do a full pass. Especially confirm the new commit-on-finish name path round-trips correctly and OT data on owned mons follows the player.
Trainer-card number-field spacing — apply the FieldLabel + RowLayout pattern to retire the last fixed-offset layouts (CardFront.qml, DefTextEdit.qml). The exact look is a human/design call.

Testing (new track — planned 2026-06-07)

Comprehensive automated test suite — full strategy in plans/testing.md. Phase 1 is now implemented under projects/tests/ (CMake/CTest harness + fixture helper; round-trip identity & money-isolation tests on the real saves; DB integrity test). ⚠️ Not yet built/run on the dev machine — next action: build in Qt Creator (or cmake --build then ctest --output-on-failure) and triage results. The first run may legitimately surface a round-trip imperfection or an offset to nudge. Then continue the phased rollout (savefile/common/db coverage → negative/integration/E2E → randomizer/fuzz → sanitizers/coverage → CI → app C++ → QML last). Remaining open questions there: coverage-gate strictness, CI host, characterizing BaseSAV.sav.
QML smoke test + first comprehensive GUI suite — BUILT 2026-06-13. tst_qml_screens (loads every screen, fails on any QML warning; main gated on it) PLUS a real-app GUI suite on the tests/helpers/guiapp.h harness: tst_gui_navigation (sweep every screen on populated + new saves), tst_gui_saveload (edit through screens → save → reopen across files, independence, randomize, byte-stability), tst_gui_input (synthesized keyboard input). All run offscreen and gate BOTH the Linux and Windows CI jobs. ✅ BUILT + RUN + GREEN on the Qt 6.11 kit 2026-06-13 (full ctest 61/61). Triage done (harness keyType/appBody StackView-class/busy-wait fixes, font allowlist, money-field objectName, detail-screen load-only smoke); a real Pokemart empty-cart crash was found + fixed (status.md Open Issues). Full writeup in status.md + testing.md.
- NEXT: continue the GUI roadmap (plans/testing.md → "Broader GUI coverage"): synthetic fixture matrix (generated edge-case saves in a clearly-synthetic folder), detail-screen flows (pokemon→select→details with a real selection — verifies the cold-load bindings the smoke test leaves load-only), keyboard shortcuts, drag & drop flows, shell/fragment smoke, Red/Blue + game-state fixtures, and ASan-under-GUI on Linux CI.

i18n (new track — added 2026-06-13)

The Qt Linguist pipeline is in place (reference/i18n.md); English ships. Open follow-ups, none urgent:

In-app language switcher — DEFERRED (decided 2026-06-13): with only en_US and no in-app Options/Settings screen yet, there's nothing to switch to. Revisit once a second locale + the Options screen exist. ui/language is the hook; it'll need engine retranslate() + re-evaluating bindings plus the Settings UI. Until then, locale = system/registry only.
Ship a real second locale when wanted — add one TS_FILES line, run update_translations in Qt Creator, translate in Linguist. Proves the path end-to-end (currently only en_US exists).
Optional: wrap the deliberately-skipped tiny format prefixes ("L"+level, "x"+count, "No."+n) as qsTr().arg() if a real locale needs them.
Out of scope / much later: game-data name localization (Pokémon/move/item) — region/encoding -bound save data, a separate effort; do NOT route through Qt translations.

Pending decisions — tracked temporary exceptions (resolve, don't let linger)

These are deliberate "dirty patches" the test pass put in to keep things working until a human makes the call. Each is a single-truth question with (likely) one correct answer.

type2 single truth (single-type Pokémon). Real saves store a single type inconsistently — sometimes 0xFF, sometimes a duplicate of type1 — so the load/expanded side officially tolerates both (byte fidelity: read it back exactly as loaded, change only when asked; isCorrected() was patched 2026-06-08 to accept either form). Undecided: the single canonical form the editor should write when it itself generates/corrects a single type (0xFF vs duplicate-of-type1). Today a generated single type collapses to type2=0xFF internally and save() writes the duplicate (type1) when type2Explicit is false. Pick the real truth, then tighten isCorrected()/save() to it. Refs: pokemonbox.cpp isCorrected()/update()/save(), reference/fix-patterns.md.

(Resolved 2026-06-08: isMinEvs() ||→&& — confirmed a bug and fixed, not a pending decision.)

Optional cleanup

Delete the now-unused menu files (name/NameDisplayMenu.qml, NameDisplayMenuNoTileset.qml, TilesetMenu.qml) and their app.qrc entries once the button-based editors are settled.
Give the inline combos (StarterEdit, Rival starter, SimulatedTilesetCombo) the same hover underline as the Select* combos, for consistency.
Maps screen — confirm map data loads and appBody.push to MapDetails works.

Don't chase

The system-wide terminated abnormally / Qt-debugger pop-ups (also in Notepad/taskbar) — environment/Qt-install issue, not this app. See status.md → "The crashes — two different things".

Longer Term

Window chrome & selective fluidity, randomization features, map editor, full screen coverage — plans/future.md. These are human/design calls.

Don't chase

The system-wide terminated abnormally / Qt-debugger pop-ups (also in Notepad/taskbar) — environment/Qt-install issue, not this app. (The in-app GC use-after-free crashes ARE fixed.) See status.md → "The crashes — two different things".

Longer Term

Window chrome & selective fluidity — an exploratory idea (lock to the design resolution but let tall-list screens like Pokemart grow; leaning toward frameless custom chrome with app-driven per-screen sizing). Direction + open questions in plans/future.md → "Window Chrome & Fluid Layout". A human/design call.
Randomization features, map editor, full screen coverage — plans/future.md.
The system-wide terminated abnormally / Qt-debugger pop-ups (also in Notepad/taskbar) — environment/Qt-install issue, not this app. (The in-app GC use-after-free crashes ARE fixed.) See status.md → "The crashes — two different things".

Cosmetic / layout (in progress)

✅ Combo box popups now scroll on long lists (s13k); ✅ hover pen tints light (s13k).
Window chrome & selective fluidity — an exploratory idea (lock to resolution but let tall-list screens like Pokemart grow; possibly borderless custom chrome + drag handle that animates back to the design size). Direction + open questions captured in plans/future.md → "Window Chrome & Fluid Layout". Suggested low-risk first step: make tall-list screens internally scrollable / height-flexible before attempting custom window chrome. A human/design call.

Longer Term

See plans/future.md — randomization features, map editor, full screen coverage.