A second rule set chosen at setup (CatanRuleSet): commodities, knights, walls,
barbarians and the event die, city improvements and metropolises, and the
54 progress cards. Everything lives behind CatanState.ck, so the base game's
state and rules are untouched. Multi-seat decisions run through one chore queue.
Also wraps three multi-part log lines so each event is one entry.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_016eTor2tpxKb4dXbd8RqnvZ
- Hexes that paid out for the turn's roll are lit (not on a 7, not the robber's hex, not before the roll).
- Compact layouts fold Road/Settlement/City into one Build button with a menu.
- Harbour markers are drawn with a resource pictogram (? for 3:1), larger, and sit a full hex radius off the coast so the piers show which vertices they serve.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_018njn2eMPAYqAoE49hGRg2X
Rules
- A road may only start from a crossing where the player has a building, or
where one of their roads ends AND no opponent's settlement or city stands.
roadConnects ignored opponents' buildings, so a road (the computer's included)
could be laid straight past a settlement. Also applies to Road Building.
- Enforce the piece supply: 15 roads, 5 settlements, 4 cities. It was
unlimited.
- Tests: building past a settlement and a city, free roads, own settlement not
blocking, and running out of each piece.
UI
- The layout rescaled the board whenever it became the player's turn, because
the action buttons only existed then. Everything outside the board now has a
fixed height: the action bar always shows every button (disabled, not absent),
the prompt is pinned to two lines, player cards always draw every line, and
the dice total is always present. Rule recorded in docs/UI-catan.md.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
Adds the Katan screen on top of the WALA-119 engine, built in separable layers
so the look, wording, layout and interaction can each be changed alone:
- CatanStyle: every colour and proportion (light/dark), via LocalCatanStyle
- CatanText: every word on screen (pure functions of state)
- CatanGeometry: lattice -> pixels and tap hit-testing (pure, unit-tested)
- CatanInteraction: what is tappable and what a tap means, derived only from the
engine's legal-move list (pure, unit-tested)
- CatanBoardView / CatanPanels / CatanDialogs: stateless composables
- CatanScreen: wiring to MatchController, portrait and landscape layouts
- CatanPreviews: Android Studio previews of real positions
- docs/UI-catan.md: how to adapt each layer, rules of the road, known gaps
Wiring: PlayScreen branch, GameSetupScreen options (3-4 players, game to
5/8/10/12), and a GameDescriptor.devPreview flag so dev builds can open Katan
while production still hides it until two-device play is verified.
Tests: CatanGeometryTest, CatanInteractionTest, CatanTextTest, CatalogTest.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
The Option 2 migration dropped Gitea's 5c797c8, which added GameCatalog.CATAN
and a Katan entry (available = false: Coming soon on dev, hidden from
production). CatanRules.kt references GameCatalog.CATAN, so the merge did not
compile without it. Taken verbatim from feature/catan-engine, with the
matching CatalogTest assertions.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
Brings in the Katan rules engine, board, model and AI (WALA-119), its 51 unit
tests and docs/RULES-catan.md. Only conflict was docs/README.md, which the
GitHub-lineage migration had dropped: kept the engine branch's index and
pointed the CI link (docs/CI.md no longer exists) at the root README.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
Runs on the Gitea runner crash intermittently on three JDK vendors. Upload
hs_err/replay logs on failure so the failing frame survives the job
container, and drop C2 (TieredStopAtLevel=1) for the Gradle and Kotlin
daemons to test whether the optimising JIT is the trigger.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01YGRgChhwbwWr5yYrHTd9pL
The Gitea runner's container image has node but no jq, so the first dev deploy
died with exit 127 in stage-pages-site.sh. Same behaviour, tested against mock
servers for: other channel published, 200-with-HTML, 404, and unreachable.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
The sheet arrives over the socket just after connected flips, so on a slower
runner the comparison raced it (ScorekeeperSessionTest, line 167). Same
pattern the sibling test already uses.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
setup-java's temurin distribution resolves versions through api.adoptium.net,
which fails TLS from the runner (EPROTO) while api.azul.com is reachable.
Any JDK 17 builds the project.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
- plgVersionRunOffset=1000: stamped build number is run number + offset, so
Gitea's restarted run counter cannot produce a versionCode below an
installed one (GitHub reached prod 30, dev ~172). Tags and versionName use
the same number.
- Drop the GitHub release steps; plg.wallacenet.ca is the only channel.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
Keeps Gitea's history reachable (also preserved as gitea-pre-github-main)
while the tree stays the GitHub one: package org.prolibertate.games, OTA
from plg.wallacenet.ca. Gitea's package rename and Gitea-releases updater
are intentionally not carried over.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
Pages serves index.html with 200 for unknown paths when there is no
404.html, so a never-published channel looked published and jq failed on
the HTML. Add a real 404 page and treat a 200 without a latest.json as
absent.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
A push-triggered first deploy cannot set the first_deploy input, and the
domain answers 403/1014 until the Pages project exists. 1014 means nothing
exists to preserve, so handle it like a 404. Other 403s still abort.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
Both workflows gain a deploy-pages job that stages their channel into one
Pages site, carrying the other channel over from the live site. The in-app
updater now reads /prod|/dev/latest.json from plg.wallacenet.ca instead of
the GitHub API. GitHub releases are still published for one transition
deployment so existing installs can fetch the new updater.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01JdMNbY5hLR7FjtfEE712xG
The island game as a pure-Kotlin state machine under game/catan/, wired to
the upstream GameRules / GameAi contract and fully unit-tested on the JVM.
- game/catan/: CatanModel (state, moves, phases), CatanBoard (19-hex
geometry, seeded island generation with the 6/8 separation rule, shared
vertex/edge graph), CatanRules (setup, production, the seven and the
robber, domestic and maritime trade, building, development cards,
Longest Road, Largest Army, win detection), CatanAi (a heuristic opponent
that reads only public information).
- GameRules.isLegalMove: a defaulted structural legality check so a game
with a combinatorial move space (Catan's trade offers) is gated the same
way as one with an enumerated move list; MatchController uses it.
- GameCatalog.CATAN: the catalogue entry, listed but not yet available
(descriptor and screen land with the screen task).
- Tests: 51 across CatanBoardTest, CatanRulesTest and CatanAiTest —
geometry invariants, generation balance, setup, production, the
seven/robber, both trade types, development-card timing, Longest Road
edge cases, Largest Army, win detection, and AI games played to a result.
- docs/RULES-catan.md, in the per-game style, with the rule choices and
the known gaps called out.
Two real engine bugs found and fixed while getting the suite green:
VERTEX_EDGES counted each shared edge once per adjacent hex, so an
interior vertex looked like it had six edges (which also double-counted
segments in the Longest Road search and duplicated setup moves); and the
applyRoll seven-transition was a discarded expression, so a seven never
left AWAIT_ROLL. Three test fixtures rolled from the PLAYING phase and two
assumed a two-seat rotation; corrected to the phases and seat order the
engine actually uses.
Verified: ./gradlew :app:testDebugUnitTest --offline — 679 tests, 0 failures.
Co-Authored-By: Paperclip <[email protected]>
Follows the Gitea repo rename: the applicationId (and its .dev variant)
now matches the repo, the launcher name and the docs, instead of carrying
the abandoned Pilgrims of Katan identity.
The JVM namespace stays org.prolibertate.games, so no source file moves;
only the installed identity changes.
Co-Authored-By: Paperclip <[email protected]>
The repo was renamed from PilgrimsOfKatan to ProLibertateGames on the
Gitea instance. Update every reference so nothing points at the old name:
- OTA updater: the repo constant and the User-Agent
- publish-release.sh: the default repo, the asset name, the release title
- CI/release workflows: the artifact names and the staged APK name
- settings.gradle.kts: the Gradle project name
- Protocol.kt / LanTransport.kt: the mDNS service label and the Wi-Fi lock tag
- docs (README, CLAUDE.md, docs/CI.md) and the update-policy test fixtures
The installed applicationId is left unchanged in this commit.
Co-Authored-By: Paperclip <[email protected]>
The bootstrap rebranded the app to Pilgrims of Katan (WALA-118 review);
the decision is to keep the upstream's own identity instead.
- Title and splash: the upstream originals restored exactly — the Wallace
tartan splash under the "Pro Libertate Games" uncial wordmark
(SplashScreen.kt, Tartan.kt), the tartan launcher icon, and the
Theme.ProLibertateGames window theme. The game screens go back to the
upstream WallaceGold/WallaceRed accents the tartan is woven from.
- Names and colours: launcher names back to "Pro Libertate Games" /
"Pro Libertate Games DEV"; the channel colours back to the clan red /
hunting green setts (resValue tartan_field and friends). The installed
identity stays ca.wallacenet.pilgrimsofkatan (+ .dev) — that is the
package, not the brand.
- Catalogue: Katan is listed under Board Games (3-4 players, original
blurb, no third-party marks) with available = false, so dev builds show
it as "Coming soon" and a production build hides it until its engine
and screen land.
- Docs: README/CLAUDE.md/docs/CI.md updated to the restored identity;
CatalogTest's invariant now names Katan as the one deliberate exception
rather than asserting nothing is unfinished.
Structural bootstrap untouched: applicationId, OTA repointed at the Gitea
releases, CI workflows, version stamping.
Verified: ./gradlew assembleDebug and testDebugUnitTest green — 628 tests,
0 failures.
Co-Authored-By: Paperclip <[email protected]>
A live publish on git.wallacenet.ca kept coming back invisible to the updater:
the create accepted "draft": false, and the release was still stored as a
draft once the APK was attached — so the anonymous list the app reads came
back empty even though every publish step reported success.
The script now attaches first, then reads the release back and PATCHes the
state if it did not stick, failing loudly if the correction is refused. A
draft is not a channel a preview build can silently become.
Also lists all 22 catalogue games in the README's table; seven were missing,
which read as "the app has 15 games".
Verified against the live instance: anonymous list shows
v1.4.1001-dev as a published prerelease with a tag-named asset, the
production /releases/latest correctly returns 404 while only a prerelease
exists, and an anonymous download through the URL the app builds matches the
built APK byte for byte.
Co-Authored-By: Paperclip <[email protected]>
Two things the first live release check on git.wallacenet.ca showed:
- The publish script uploaded the APK under whatever the build called it
(app-debug.apk), which is the name the updater only finds by its fallback
rule. The script now names the asset after its tag, so the primary rule —
"the asset whose name contains the tag" — is the one that fires.
- The instance answers for two host names and reports asset URLs on the one
it treats as canonical, so the app would have fetched its release list from
git.wallacenet.ca and its APK from elsewhere. onReportedHost() now keeps the
path the server chose and puts the download on the host the app already
trusts, tested in UpdatePolicyTest.
Co-Authored-By: Paperclip <[email protected]>
Import of ProLibertateGames (upstream main, GPL-3.0) rebranded for
Wallacenet and repointed at git.wallacenet.ca.
- Identity: applicationId ca.wallacenet.pilgrimsofkatan (+ .dev for the dev
channel), launcher names "Pilgrims of Katan" / "Pilgrims of Katan DEV",
neutral hexagon branding replacing the Wallace tartan. Uncial Antiqua keeps
its OFL licence, and the GPL-3.0 LICENSE plus upstream attribution stay.
- OTA: Updater.kt now reads this repo's own Gitea releases
(/api/v1/repos/devadmin/PilgrimsOfKatan/releases...) and BuildConfig.
OTA_ENABLED is on; Gitea answers the same JSON shape the code already parsed.
- Build numbers: versionCode = CI run number + plgVersionRunOffset (1000),
with the tag and versionName derived from the same figure. The offset is
permanent, because Gitea Actions restarts run numbers at 1 and a reset would
otherwise make Android reject the next build as a downgrade.
- CI: ci.yml publishes a dev prerelease from dev, release.yml a production
release from main, both through scripts/publish-release.sh (Gitea REST API,
runnable by hand until an Actions runner is registered).
- Docs: docs/CI.md records the channel, numbering and runner decisions;
docs/README.md seeds the per-game RULES-* conventions; CLAUDE.md carries the
agent ground rules.
Verified: ./gradlew assembleDebug and testDebugUnitTest green on JDK 17 /
compileSdk 34, both variants stamped, installing side by side under the new
identity.
Co-Authored-By: Paperclip <[email protected]>
Tally sheets let any player score at any moment (+1, -1 or a custom amount)
with a log of entries that can be taken back. Entries are stored as one-player
rows so totals, undo and sharing work unchanged.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A guest was timing out with nothing ever shown on the host. Without a
device to look at, this makes the next test say where it stops, and fixes
what was found on the way.
- The host's Sharing panel shows links received and join requests read, which
tells "nobody reached this phone" from "reached it and not understood".
- The guest says when its link is up and it is waiting for an answer.
- A request waiting shows on the host's top bar ("Sharing . 1 waiting"), and
the Sharing panel steps aside while one is up so two dialogs never stack.
- Join is a top-bar action on a sheet already under way; it was only offered
on an empty one.
- LanTransport: a host being cleaned up no longer unregisters the
advertisement or clears the address of a newer host started in its place.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01CBG7URddsUZjQ6KE1Kqf6G
A host can share its score sheet and others nearby can join it, found the
same way games are, or by typed address over a hotspot.
- A joining guest is not let in on its own say-so. The host is prompted with
the guest's name and answers Allow or Deny. A device already let in that
drops out (locked phone) is let straight back; the host can remove anyone.
- The host's sheet is authoritative. What travels is what was done (SheetOp:
add round, rename, move column, ...), not whole sheets, so changes made at
the same moment all land. A guest's own changes show at once and are laid
over the host's sheet until confirmed (SheetReplica).
- Sync asks the host for the sheet as it stands, and runs on its own when
the app returns to the front, as it does for a game table.
- Hello carries a purpose and hosts advertise one, so a game join screen does
not list score sheets and a game guest typing a sheet's address is told so.
- The Scorekeeper screen now expresses every edit as a SheetOp, so one grid
serves a local sheet, a hosted one and a joined one.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01CBG7URddsUZjQ6KE1Kqf6G
Claude
devadmin