Commit 408097e ('feat: move func to calc package') moved a helper
into pkg/calc and made pkg/util/map.go import galaxy/calc, but the
local-dev backend / gateway Dockerfiles never picked up the new
module. The synthesised go.work has no replace directive for
galaxy/calc and the build context never copies pkg/calc, so any
backend / gateway image rebuild fails with
galaxy/calc@v0.0.0: malformed module path "galaxy/calc": missing
dot in first path element
Add the missing COPY, the matching `use ./pkg/calc` line, and the
`galaxy/calc v0.0.0 => ./pkg/calc` replace to both local-dev
Dockerfiles. The local-dev stack now rebuilds cleanly and the
auto-heal flow (prune-broken-engines + pre-bootstrap reconciler
tick) finishes by spawning a fresh engine container for the new
sandbox game.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
`make rebuild` runs `compose build --no-cache backend gateway` plus
a fresh `up -d --wait`. It must therefore also reap any engine
container whose bind-mount source went away during host downtime,
otherwise the new backend image boots into a stack with the same
orphan that triggered the heal flow in the first place.
Also extend the troubleshooting note: pulling the heal-cycle fix
requires one explicit `make rebuild` so the backend image picks up
the pre-bootstrap reconciler tick. Without that, `make up` runs
the new Makefile target but the legacy backend cannot follow
through, and the developer is left staring at a `cancelled`
sandbox with no running replacement.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
After a host reboot macOS clears /private/tmp, so the per-game
bind-mount source under /tmp/galaxy-game-state/<uuid> vanishes and
Docker refuses to restart the long-lived engine container under
`restart: unless-stopped`. The container then sits in `exited` state
and the dev sandbox is unreachable until the developer manually rms
it and runs `make up` twice.
Fix `make -C tools/local-dev up` to heal this in one cycle:
1. `prune-broken-engines` (new make target wired into `up`) walks
every container labelled `galaxy-game-engine` and removes the ones
not in `running` / `restarting` state. Healthy long-lived
containers survive normal up/down cycles untouched.
2. The backend now runs a single reconciliation pass before the
dev-sandbox bootstrap (`Reconciler().Tick(ctx)` in main.go).
Without it, bootstrap would reuse the soon-to-be-cancelled game
that the periodic ticker is about to mark `removed`. The pre-tick
cascades the orphan runtime row through markRemoved → lobby
cancel before bootstrap purges terminal sandbox games and creates
a fresh one — so a single `make up` lands a working sandbox with
a brand new state directory.
README troubleshooting section documents the symptom and the
recovery so the bind-mount-source error message is greppable.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Fixes a black-canvas regression on /map after creating a science in
DEV: when Vite hot-reloads the decoder bump that adds the
`localScience` field, the live in-memory `gameState.report` keeps
its older shape with no such field, so the overlay's
`[...report.localScience]` throws inside the reactive getter and
silently aborts the map view's `$effect`. The fix wraps the spread
and the final return in `?? []` defaults — and matches the
ship-class branches for symmetry — so the overlay stays well-defined
for any partial report shape upstream consumers may carry across an
HMR boundary.
Also adds order-overlay regression tests covering the createScience /
removeScience branches plus the explicit HMR-stale shape, and a
Playwright e2e (sciences-map-regress.spec.ts) replaying the
user-reported flow: /map → /designer/science → save → /map, asserting
no map-mount-error overlay and no console errors.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Lights up the player-defined sciences feature: a table view with sort
and filter, a designer with four percent inputs and a strict
sum-equals-100 gate, and a Research-sub-row integration so the
planet production picker lists the user's sciences alongside the
four tech buttons. Phase 21 decisions are baked back into ui/PLAN.md
(no UpdateScience on the wire — write-once via createScience +
removeScience; percentages instead of fractions; sciences live under
the existing Research segment).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Send joins Modernize / Dismantle / Transfer as a lockable command:
once any of the four lands in the draft for a group, every action
button on its inspector is disabled with a "command pending"
tooltip and the banner names the queued kind. Load / Unload /
Split / Join Fleet stay non-locking — they stack legitimately on
the engine side.
Two dashed overlays now run alongside the cargo-route arrows:
- Yellow dashed track for own in-space groups, drawn from the
origin planet to the destination (matches the in-space point
colour so eye reads both as one entity).
- Green dashed track for every wire-valid sendShipGroup command
in the order draft, drawn from the source group's orbit planet
to the chosen destination. Disappears when the command is
removed from the order tab, when the engine rejects it, or
when the group has left orbit (in-space track replaces it).
Both tracks are wrap-aware via torusShortestDelta and never
participate in hit-test.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Send no longer carries a destination control inside the form: a
click on the action drops the inspector straight into map-pick
mode, and the form (ship count + confirm) only mounts after the
player chooses a destination. Cancelling the picker leaves no
form behind.
A queued Modernize / Dismantle / Transfer for a given group
locks every action button on its inspector and surfaces a banner
that points the player at the order list. Cancelling the queued
entry from the order tab releases the lock on the next render —
the derivation watches draft.commands directly. Send / Load /
Unload / Split / Join Fleet do not lock; Send is naturally
followed by an out-of-orbit state at turn cutoff, the rest can
stack legitimately.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Eight ship-group operations land on the inspector behind a single
inline-form panel: split, send, load, unload, modernize, dismantle,
transfer, join fleet. Each action either appends a typed command to
the local order draft or surfaces a tooltip explaining the
disabled state. Partial-ship operations emit an implicit
breakShipGroup command before the targeted action so the engine
sees a clean (Break, Action) pair on the wire.
`pkg/calc.BlockUpgradeCost` migrates from
`game/internal/controller/ship_group_upgrade.go` so the calc
bridge can wrap a pure pkg/calc formula; the controller now
imports it. The bridge surfaces the function as
`core.blockUpgradeCost`, which the inspector calls once per ship
block to render the modernize cost preview.
`GameReport.otherRaces` is decoded from the report's player block
(non-extinct, ≠ self) and feeds the transfer-to-race picker. The
planet inspector's stationed-ship rows become clickable for own
groups so the actions panel is reachable from the standard click
flow (the renderer continues to hide on-planet groups).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Two follow-up fixes after the initial Phase 19 landing:
1. The IncomingGroup dashed trajectory was drawn between raw
(x1, y1) and (x2, y2) world coordinates. On torus wrap mode
this took the long way around when origin and destination
sat near opposite seams. The line now picks endpoints via
`torusShortestDelta` so the segment crosses the seam when
that's the shorter visual path. The interpolated clickable
point follows the same unwrapped vector. The same helper
fixes the in-hyperspace position for local / foreign groups.
2. On-planet local and foreign groups previously rendered as
small offset points next to every populated planet, which
turned the canvas into noise as soon as a player held more
than a handful of planets. The map no longer renders any
in-orbit group; the planet inspector grows a compact
"stationed ship groups" subsection
(`lib/inspectors/planet/ship-groups.svelte`) that lists
each in-orbit group as a row of `<race> · <class> · <count>
ships · <mass>`. Race attribution: LocalGroup → the player's
race, OtherGroup on a foreign-owned planet → the planet's
owner, OtherGroup elsewhere → "foreign" placeholder. Rows
are non-interactive in Phase 19; Phase 21+ will deep-link
into the ship-groups table view with a (planet, race) filter.
Tests:
- `state-binding-groups.test.ts` swaps the on-planet rendering
expectation for the new "no map primitive" rule, and adds a
regression that asserts the incoming line crosses the torus
seam via `torusShortestDelta`.
- new `inspector-planet-ship-groups.test.ts` covers row
composition, the destination-mismatch filter, the
in-hyperspace exclusion, the foreign-planet owner fallback,
and the empty-state collapse.
- `inspector-planet.test.ts` and `inspector-ship-group.spec.ts`
pick up the new prop chain (`localShipGroups`,
`otherShipGroups`, `localRace`).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The pre-existing `Deltas` helper used the height to wrap the x-axis,
which silently produced wrong values on any rectangular galaxy
(`w != h`). Square galaxies — the only configuration the engine
ships today — masked the bug, so it stayed in tree.
`Deltas` is now a thin wrapper around the new `ShortestDelta(a, b,
size)`, which returns the signed per-axis shortest delta on a 1-D
circle (range `(-size/2, size/2]`). The signed flavour is what the
Phase 19 ship-group renderer needs to draw an IncomingGroup
trajectory across the torus seam; `Deltas` continues to return the
pair of absolute deltas for distance computation.
Adds `pkg/calc/map_test.go` with table-driven coverage for both
helpers, including a regression that exercises the rectangular
case the bug was hiding behind, and the half-circumference
tie-break.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The root +layout.svelte redirects anonymous traffic to /login, so a
fresh CI browser context never gets to render the lobby's
DEV-gated synthetic-report section — the previous spec relied on
leftover session state in the local browser and silently broke on
clean runners (local-ci run 23).
Bootstrap the session through /__debug/store before navigating to
/lobby: load a device keypair, set a deterministic device session
id. The synthetic flow itself still bypasses the gateway entirely;
the seed only ensures `session.status === "authenticated"` so the
layout guard lets the lobby through.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Closes Phase 19's UI surface. The inspector dispatches on the
selection variant: local / other groups render class, count, the
four tech levels, mass, cargo (type + amount when loaded),
location (planet name on-orbit, from/to/distance in hyperspace),
and — for local groups only — fleet membership + state. Incoming
groups surface origin / destination / distance / speed and the
inline ETA = ceil(distance / speed); zero speed collapses to the
designer's existing "—" placeholder. Unidentified groups render
just the (x, y) coordinates and the no-data hint, mirroring the
unidentified planet treatment.
Layout / inspector-tab plumbing:
- inspector-tab.svelte derives selectedShipGroup against the
rendered report and mounts <ShipGroup /> when the planet
branch doesn't match. Stale refs (an index that no longer
resolves after a turn refresh) collapse cleanly to the empty
state.
- +layout.svelte mounts <ShipGroupSheet /> alongside the
existing planet sheet on mobile; both share the
`effectiveTool === "map"` guard and clear-on-close.
i18n: en + ru both grow ~30 keys under
`game.inspector.ship_group.*`. Adding a key to one without the
other is a TS error (TranslationKey is `keyof typeof en`), so the
Russian mirror stays mandatory.
Tests:
- inspector-ship-group.test.ts exercises every variant —
on-planet local, in-hyperspace local, cargo-loaded local,
foreign, incoming with ETA, incoming with zero speed,
unidentified, plus the missing-planet `#NN` fallback.
- tests/e2e/inspector-ship-group.spec.ts is a smoke spec that
drives the DEV-only synthetic-report loader from /lobby
through navigation to /games/synthetic-XXX/map.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Wires Phase 19's data and rendering layers without yet adding the
inspector UI:
- game-state.ts grows ReportLocalShipGroup / ReportOtherShipGroup
/ ReportIncomingShipGroup / ReportUnidentifiedShipGroup /
ReportLocalFleet types and walks the matching FlatBuffers
vectors (LocalGroup, OtherGroup, IncomingGroup,
UnidentifiedGroup, LocalFleet) inside decodeReport. The Tech
map is folded into the fixed-shape ShipGroupTech struct;
cargo strings normalise to the closed CargoLoadType | "NONE"
union; UUIDs come back as canonical 36-char strings.
- synthetic-report.ts mirrors the new fields so the DEV-only
lobby loader can feed JSON produced by legacy-report-to-json
straight into the live UI surface.
- selection.svelte.ts widens its discriminated union with a
`kind: "shipGroup"` branch carrying a ShipGroupRef
(local UUID / other / incoming / unidentified by index).
- world.ts adds Style.strokeDashPx and render.ts.drawLine
honours it via manual segmentation (PixiJS v8 has no native
dash API). Ignored on points and circles.
- state-binding.ts now returns { world, hitLookup }: the
hit-lookup map keys every primitive id back to a concrete
HitTarget so the click handler can dispatch to selectPlanet
or selectShipGroup. Ship-group primitives live in a separate
ship-groups.ts that emits one point per local / other /
unidentified group, plus a dashed origin→destination line +
clickable point per incoming group. Position is interpolated
along the trajectory for in-hyperspace groups.
- map.svelte threads the hitLookup into handleMapClick.
Vitest:
- tests/helpers/empty-ship-groups.ts exposes EMPTY_SHIP_GROUPS
so existing fixtures can spread the new five empty arrays
without enumerating every field.
- state-binding-groups.test.ts covers each group variant's
primitive geometry and lookup correctness.
- All previously-existing fixture builders pick up the spread
so GameReport stays a complete object.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The parity rule from ui/PLAN.md says every UI phase that decodes a
new Report field must extend the legacy converter in lockstep.
Phase 19 brings ship groups (LocalGroup / OtherGroup /
UnidentifiedGroup / IncomingGroup) and LocalFleet onto the wire-
compatible UI surface; this commit teaches
tools/local-dev/legacy-report to populate the three sections that
exist in the legacy text format:
- "Your Groups" → []LocalGroup. Cargo type, load, fleet name,
state, on-planet vs hyperspace position (origin / range) all
decoded; LocalGroup.ID is synthesised deterministically from
the per-report group index so re-running the converter
produces byte-identical JSON. Speed is left zero — the legacy
table doesn't expose it.
- "Your Fleets" → []LocalFleet. Origin / range / state mirror
the row layout used by Killer / Tancordia variants; gplus's
state-less rows still resolve.
- "Incoming Groups" → []IncomingGroup. Origin / destination
names — and `#NN` by-id references — resolve against the
parsed planet tables. Because the section can land before
"Your Planets" in some engines, group / fleet / incoming rows
are buffered and resolved in `parser.finish` after every
planet is known.
Battles, OtherGroup (only ever in battle rosters), and
UnidentifiedGroup stay out of scope — README.md spells out what
remains not-derivable.
Adds Killer031–033 / TSERCON_Z032–033 / Tancordia036–039 fixtures
to the dg directory and exercises three of them through new
TestParseDg{Killer031,Tancordia037,KNNTS041} smoke tests, plus
inline tests for each new section parser. Drops the stale
KNNTS039.json artefact left over from Phase 18 development.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Locks in the synthetic-report parity rule as a global "Assumptions
and Defaults" entry in ui/PLAN.md: every phase that extends the
server->UI report contract must also extend the legacy parser in
the same PR (or document in tools/local-dev/legacy-report/README.md
why the new field cannot be derived from legacy text). The Go side
already enforces shape compatibility via the pkg/model/report
import; this rule extends that mechanical guard to "did we remember
to wire the new field through".
ui/docs/testing.md grows a "Synthetic reports for visual testing"
section with the full conversion -> load -> compose loop and the
two operational gotchas (no network on synthetic ids, page reload
clears the in-memory map).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds api/synthetic-report.ts, an in-memory registry + JSON->GameReport
decoder for synthetic-mode game sessions. The lobby grows a
import.meta.env.DEV-gated "Synthetic test reports" section with a
JSON file picker; loading a file registers the decoded report under
a synthetic-<uuid> id and navigates to /games/<id>/map.
The in-game shell layout detects the synthetic id range, takes the
report straight from the registry via gameState.initSynthetic, and
deliberately skips both galaxyClient.set and orderDraft.bindClient.
Order auto-sync stays silent: scheduleSync already short-circuits on
non-UUID game ids, and without a bound client the network path is
unreachable. applyOrderOverlay continues to project locally-valid
draft commands onto the rendered report so renames / production
choices / route edits are visible immediately.
A page reload loses the in-memory entry and redirects to /lobby —
synthetic mode is a debug affordance, not a session.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
A Go module under tools/local-dev/legacy-report that converts the
"dg" / "gplus" engine .REP files in tools/local-dev/reports/ into the
JSON shape of pkg/model/report.Report. The output drives a DEV-only
synthetic-mode loader on the UI lobby so the map, inspectors, and
order-overlay can be exercised against rich game states without
playing many turns end-to-end.
Scope is intentionally narrow: only the fields the UI client decodes
today (planets, players, own ship classes, header). Importing
pkg/model/report keeps the parser and the typed contract in lockstep
— any backwards-incompatible schema change breaks the tool's
compilation before it ships. The README spells out the parity rule
for extending the parser alongside future UI decoders.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Wires pkg/calc/ship.go into the WASM Core boundary as seven thin
wrappers (DriveEffective, EmptyMass, WeaponsBlockMass, FullMass,
Speed, CargoCapacity, CarryingMass). The ship-class designer reads
Core through a new CORE_CONTEXT_KEY populated by the in-game layout
and renders a five-row preview pane (mass, full-load mass, max
speed, range at full load, cargo capacity) that updates reactively
on every form edit and on the player's localPlayer{Drive,Weapons,
Shields,Cargo} tech levels — three of which are now decoded from
the report's Player block alongside the existing localPlayerDrive.
CarryingMass is the seventh wrapper added to the original six-function
list so that "full-load mass" composes through pkg/calc/ functions
without putting math in TypeScript.
Mirrors the validateEntityName MAX_LENGTH on the form input so the
field stops accepting characters once the limit is hit. The
validator still runs and surfaces the localised reason if a paste
overshoots; the maxlength is purely a typing-time guardrail.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds an explicit acceptance line documenting that pending Save /
Delete actions reflect on the table immediately via
applyOrderOverlay (already exercised by the new Vitest + Playwright
suites). Touches a watched path so the local-ci runner picks up the
retrigger after the previous flaky run.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Run 19 failed on a known flake in pkg/util.TestRandomSuffixGenerator
(line 289 — generator can produce the same 4-char suffix in two
consecutive draws out of 10000). Empty commit retriggers the
runner; flake is unrelated to Phase 17.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Phase 17 lights up the ship-class table and designer active views,
extends the order-draft pipeline with createShipClass and
removeShipClass commands, and projects pending Save/Delete actions
through applyOrderOverlay so the table reflects the player's
intent before auto-sync lands. The plan is corrected in the same
patch: per game/rules.txt, ship classes are designed once and
cannot be edited — the engine has no Update command, so the UI
exposes only Create + Delete.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The cargo-route picker filtered out unidentified planets, so an
early-game player who had spotted but not surveyed a destination
could not configure a route to it — the engine has no such
restriction (`game/internal/controller/route.go.PlanetRouteSet`
only checks ownership of the origin and `util.ShortDistance(...) <=
FligthDistance`). Drop the unidentified guard and document the
contract in `cargo-routes-ux.md` plus a comment over `reachableSet()`.
Pick-mode dim now drops both alpha and tint on out-of-reach
planets so bright shapes (`STYLE_LOCAL` is `0x6dd2ff`) collapse
into a single muted gray. The single-channel `dimAlpha=0.3` was too
gentle against the dark theme — the user reported the dim wasn't
visible. Tighten to `dimAlpha=0.35 + dimTint=0x303841`; restore
both on tear-down.
Also threads through the user's `pkg/calc/race.go.FligthDistance`
addition: `calc-bridge.md` records the new Go-side reference (the
engine's `Race.FlightDistance()` already wraps it), and the picker
comment points at the canonical formula location.
Tests:
- `inspector-planet-cargo-routes.test.ts` adds two cases — a
reach-spans-every-kind case (own + foreign + uninhabited +
unidentified all picked when in range) and a successful pick to
an unidentified destination.
- All 356 vitest cases + chromium-desktop / webkit-desktop e2e
cargo-routes pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add per-planet cargo routes (COL/CAP/MAT/EMP) to the inspector with
a renderer-driven destination picker (faded out-of-reach planets,
cursor-line anchor, hover-highlight) and per-route arrows on the
map. The pick-mode primitives are exposed via `MapPickService` so
ship-group dispatch in Phase 19/20 can reuse the same surface.
Pass A — generic map foundation:
- hit-test now sizes the click zone to `pointRadiusPx + slopPx` so
the visible disc is always part of the target.
- `RendererHandle` gains `onPointerMove`, `onHoverChange`,
`setPickMode`, `getPickState`, `getPrimitiveAlpha`,
`setExtraPrimitives`, `getPrimitives`. The click dispatcher is
centralised: pick-mode swallows clicks atomically so the standard
selection consumers do not race against teardown.
- `MapPickService` (`lib/map-pick.svelte.ts`) wraps the renderer
contract in a promise-shaped `pick(...)`. The in-game shell
layout owns the service so sidebar and bottom-sheet inspectors
see the same instance.
- Debug-surface registry exposes `getMapPrimitives`,
`getMapPickState`, `getMapCamera` to e2e specs without spawning a
separate debug page after navigation.
Pass B — cargo-route feature:
- `CargoLoadType`, `setCargoRoute`, `removeCargoRoute` typed
variants with `(source, loadType)` collapse rule on the order
draft; round-trip through the FBS encoder/decoder.
- `GameReport` decodes `routes` and the local player's drive tech
for the inline reach formula (40 × drive). `applyOrderOverlay`
upserts/drops route entries for valid/submitting/applied
commands.
- `lib/inspectors/planet/cargo-routes.svelte` renders the
four-slot section. `Add` / `Edit` call `MapPickService.pick`,
`Remove` emits `removeCargoRoute`.
- `map/cargo-routes.ts` builds shaft + arrowhead primitives per
cargo type; the map view pushes them through
`setExtraPrimitives` so the renderer never re-inits Pixi on
route mutations (Pixi 8 doesn't support that on a reused
canvas).
Docs:
- `docs/cargo-routes-ux.md` covers engine semantics + UI map.
- `docs/renderer.md` documents pick mode and the debug surface.
- `docs/calc-bridge.md` records the Phase 16 reach waiver.
- `PLAN.md` rewrites Phase 16 to reflect the foundation + feature
split and the decisions baked in (map-driven picker, inline
reach, optimistic overlay via `setExtraPrimitives`).
Tests:
- `tests/map-pick-mode.test.ts` — pure overlay-spec helper.
- `tests/map-cargo-routes.test.ts` — `buildCargoRouteLines`.
- `tests/inspector-planet-cargo-routes.test.ts` — slot rendering,
picker invocation, collapse, cancel, remove.
- Extensions to `order-draft`, `submit`, `order-load`,
`order-overlay`, `state-binding`, `inspector-planet`,
`inspector-overlay`, `game-shell-sidebar`, `game-shell-header`.
- `tests/e2e/cargo-routes.spec.ts` — Playwright happy path: add
COL, add CAP, remove COL, asserting both the inspector and the
arrow count via `__galaxyDebug.getMapPrimitives()`.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The desktop spec previously asserted the read-only `inspector-planet-
field-production` row for an owned planet. Phase 15 replaced that
row with the interactive production component on the local-planet
branch — the assertion now confirms the component is mounted and
the legacy field is absent.
Adds the second end-to-end command (`setProductionType`) with a
collapse-by-`planetNumber` rule on the order draft, the segmented
production-controls component on the planet inspector, the FBS
encoder/decoder pair for `CommandPlanetProduce`, and the
`localShipClass` projection on `GameReport`. Forecast number is
deferred and tracked in the new `ui/docs/calc-bridge.md`.
Phase 14's auto-sync calls `uuidToHiLo` on every layout boot. The
existing Phase 10 e2e specs use a placeholder string `test-shell`
as the game id, which throws in the FBS request encoder and
surfaced as a noisy `console.warn` plus a flaky webkit-desktop
test on the local-ci ARM runner.
`OrderDraftStore.hydrateFromServer` and `scheduleSync` now skip
when the active game id isn't a real UUID — the auto-sync path is
inert for fixture data and the placeholder-warning is gone. The
Phase 10 spec switches to a deterministic UUID
(`10101010-1010-1010-1010-101010101010`) so future Phase 14+
specs don't have to special-case it.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Previously a cancelled / finished / start_failed sandbox game would
hang in the dev user's lobby until manually cleaned up — `make up`
would create a new running game alongside it but the dead tiles
piled up. Now backend's `devsandbox.Bootstrap` deletes every
terminal sandbox game owned by the dev user before find-or-create
runs, so the lobby always shows exactly one running tile.
Schema: `runtime_records` and `player_mappings` gain
`ON DELETE CASCADE` on their `game_id` foreign keys so a single
`DELETE FROM games` cleans every referencing row in one write.
Pre-prod migration rule applies — change goes into
`00001_init.sql`, not a new migration.
API: `lobby.Service.DeleteGame` is the new destructive helper that
backs the bootstrap purge. It bypasses the cancel-cascade-notify
pipeline; production callers must stay on the regular lifecycle.
The dev-sandbox docs in `tools/local-dev/README.md` spell out the
new behaviour.
Tests:
- backend/internal/lobby/lobby_e2e_test.go gains
`TestDeleteGameCascadesEverything` proving CASCADE works
end-to-end against a real Postgres testcontainer.
- backend/internal/devsandbox keeps its existing terminal-status
contract test; the new `purgeTerminalSandboxGames` helper rides
on the same `terminalSandboxStatus` predicate.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces the manual Submit button with an auto-sync pipeline driven
by `OrderDraftStore`: every successful add / remove / move
coalesces a `submitOrder` call so the engine always mirrors the
local draft. Removing the last command sends an empty cmd[] PUT —
the engine, repo, and rest model now accept that as a valid
"player cleared their draft" state.
`hydrateFromServer` is now invoked unconditionally on game boot so
a fresh device picks up the player's stored order, and the local
cache is overwritten by the server's view (server is the source of
truth).
Header replaces the static "race ?" + turn counter with a single
headline string `<race> @ <game>, turn <n>`, sourced from the
engine's Report.race + the lobby's GameSummary.gameName + the live
turn number, with a `?` fallback while any piece is loading.
Tests:
- engine: empty PUT round-trips, repo round-trips empty Commands
- order-draft: auto-sync sends full draft on every mutation,
rejected response surfaces error sync status, rapid mutations
coalesce, server hydration overwrites cache
- order-tab: per-row status flips through the auto-sync lifecycle,
remove → empty cmd[] PUT, rejected → retry button
- inspector overlay: applied + valid + submitting all participate
in the optimistic projection
- header: live race / game / turn rendering with fall-back
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The Phase 14 follow-up surfaced a footgun: `make up` reuses any
pre-built backend / gateway images and silently misses route table
or transcoder edits. Add a dedicated section to the README that
points at `make rebuild`, plus the engine-only path through
`make stop-engines` + `docker rmi`.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The owner reported two symptoms after pulling the Phase 14 stack:
1. user.games.order.get answered with `unimplemented: message_type
is not routed`. The gateway/backend code was correct, but the
local-dev compose images were stale — `make rebuild` picked up
the new routes table and the symptom went away. To prevent this
class of regression from depending on docker-image freshness,
gateway/internal/backendclient/routes_test.go now asserts that
every authenticated MessageType constant declared in
pkg/model/{user,lobby,order,report} is registered, and verifies
that user.games.order.get specifically resolves to the game
command client.
2. The inspector kept the un-renamed name after a successful submit.
ui/frontend/tests/inspector-overlay.test.ts mounts the inspector
tab against a real OrderDraftStore + a stubbed GameStateStore
and walks the full happy path (add planetRename → markSubmitting
→ applied → simulate refresh) plus the integration scenario
driven through the order-tab Submit button. Both cases pass —
the underlying overlay path is reactive and resilient to a
refresh that returns the un-renamed snapshot. The original
in-browser symptom was the rebuilt-image freshness issue from
point 1; this test pins the reactive contract for future
refactors.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Wires the first end-to-end command through the full pipeline:
inspector rename action → local order draft → user.games.order
submit → optimistic overlay on map / inspector → server hydration
on cache miss via the new user.games.order.get message type.
Backend: GET /api/v1/user/games/{id}/orders forwards to engine
GET /api/v1/order. Gateway parses the engine PUT response into the
extended UserGamesOrderResponse FBS envelope and adds
executeUserGamesOrderGet for the read-back path. Frontend ports
ValidateTypeName to TS, lands the inline rename editor + Submit
button, and exposes a renderedReport context so consumers see the
overlay-applied snapshot.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Plumbs the map → inspector pathway: a click on a planet selects it
through the new SelectionStore, the sidebar Inspector tab swaps
its empty-state copy for a per-kind read-only field set, and a
mobile-only bottom-sheet mirrors the same content over the map.
Field projection in api/game-state.ts now surfaces every documented
planet field.
Ilia Denisov