Connect revealed Harass force allocation and exact casualties

Harass & Withdraw now admits legal Basic/Advanced mixed ordinary/elite commitments and multi-sector returns. A shared physical quote validates every returned counter against one exact legal dial/support complement. After both plans are revealed, the owner selects an ambiguous allocation; unique returns remain automatic. Selected counters become public before Traitor declarations, while physical transfer waits for atomic settlement and opposing-Traitor cancellation. This information is never added to a sealed Battle Plan or complete-plan inspection. Existing optional-module and Richese/Ecaz-allied gates remain.

The player panel, four AI profiles and saved event use the same quote. A versioned receipt binds the revealed plans, current typed location pools, battle and owner. Missing or altered receipts fail before reads, normalization and writes; concurrent choices select once. Previously valid legacy unique plans continue without reopening. Winner casualties now equal the selected dialed group, preventing a second smaller physical commitment after its undialed complement has returned.

Independent source and implementation review passed. Pure, engine, control and SQLite regressions cover typed/sector legality, exact complement, private pre-reveal views, public selected views, stale/malformed state, cross-owner actions, raced allocations, selected restoration, automatic exact casualties, opposing-Traitor cancellation and mandatory disposal. Types, lint and 4850 offline tests, production build and 43 HTTP/session tests pass. The first broad check caught the exhaustive decision registry and test union-narrowing omissions; those were fixed before the final stable checks. Browser testing corrected the incomplete-allocation selector label before verification.

Eight genuine card-variant Basic/Advanced games across all four homogeneous AI profiles completed with 9956 accepted actions, 266 periodic JSON restores and zero rejected candidates. Three Harass plans/returns and two Recruits plays occurred; no ambiguous allocation arose in these seeds. The new choice is exercised by targeted engine, authenticated recovery and browser cases, not inferred from those full-game samples. Full variant and combined-expansion acceptance remains incomplete.

A fresh isolated browser room used genuine Advanced card-variant setup followed by a conserved two-sector battle fixture. Actual controls rejected a partial return, selected two ordinary counters from sector nine, and preserved the selected state after refresh. Following actual owner and engine-driven opponent Traitor declines, two counters returned, the exact remaining ordinary/elite pair went to the Tanks, support cost one, and Harass discarded once. Final refresh restored 18 reserves, two Tanks, 29 spice and the same private Traitor. This is targeted browser evidence, not a complete human game.

All 1038 opening games and original seat/recovery/entry/handover records are unchanged; 1072 rooms remain including new QA. SQLite integrity and foreign keys pass. The healthy development server was reused without restart or reset. No mode/publication gate was opened.

Verified source tree: 61ee78f6e2bce47cb322b04f896db27fef28c56ba2097571a1d85bad5c121156
This commit is contained in:
Quintine committed 2026-09-14 06:37:23 +10:00
1 parent 640342527a
commit 62e87d2b96
18 files changed
+1654 -71

No files matched your search

+13 -14
View File
@@ -6,19 +6,19 @@ gated.** Prototypes do not open completion or publication gates.
## Current checkpoint and work
Latest verified checkpoint: `368a7a5`, [reusable game samples](VERIFICATION_WORKFLOW.md#reusable-base-and-faction-sample-games).
Types, lint and **4,823 offline tests** passed. Ten genuine base samples and
six expansion samples completed in 6,746 accepted actions with 175 periodic
save/restore checks. Base samples had no rejected candidates; expansion samples
retained 82 guarded arrival candidates before legal alternatives. The tooling
checkpoint changed no application, dependency, HTTP or database code.
All 1,038 opening games and original seat/recovery records were unchanged.
Latest verified checkpoint: `6403425`, [AI arrival preflight](BOT_ARRIVAL_PREFLIGHT.md).
Types, lint and **4,829 offline tests** passed. Eighteen genuine base and expansion
samples completed in 9,470 accepted actions with 247 periodic save/restore checks
and no rejected candidates. Accepted gameplay matched the prior same-seed runs
apart from random event identifiers. All 1,038 opening games and original
seat/recovery records were unchanged; the browser test seat reconnected.
Earlier evidence remains in the [implementation history](IMPLEMENTATION_STATUS.md).
The connected [Harass & Withdraw prototype](HARASS_WITHDRAW_RUNTIME.md) supports either
battle-card slot, private category inspections, exact undialed normal/elite
returns, opponent-Traitor cancellation, mandatory disposal, controls and all
four AI profiles. Physical allocation must be unique. Richese card combinations
four AI profiles. New battles also have [revealed physical allocation](HARASS_ALLOCATION.md) for
ordinary/elite mixtures and multi-sector returns. Richese card combinations
and additional optional modules stay gated. Withdrawal before explosion and
card-specific discard precedence are explicitly labeled inferences. The Stone
Burner timing question is pending; no user answer is assumed. Reinforcements
@@ -30,12 +30,11 @@ and foreign-key checks passed, and refresh restored the test seat. No restart,
database reset or recurring automation was needed. Exact verification, source
fingerprints and delivery are recorded in the private report and Git message.
Current work: [AI arrival preflight](BOT_ARRIVAL_PREFLIGHT.md) reuses the engine's
existing reaction quote to avoid repeatedly proposing guarded ordinary entries.
The adapter preserves legal single reactions, faction exemptions, contribution
routing and private information; specialized routes retain their own adapters.
The Moritani Secret Ally audit retains the shared pending hidden-reaction UX
question. No arrival-order ruling or module/publication gate is changed.
Current work: [Harass force allocation](HARASS_ALLOCATION.md) connects the missing
player choice after reveal, with shared legal AI, exact winner losses and saved
continuation. Unique returns stay automatic and sealed plans stay unchanged.
The remaining Leader Skill bands retain their existing pending rulings; this
checkpoint introduces no new arrival-order or hidden-reaction policy.
## Authoritative navigation