Files
SVSimServer/SVSim.BattleEngine.Tests/SessionEngine/CaptureReplayReshuffleRootCauseTests.cs
gamer147 addeb021d2 fix(battlenode): shadow engine tracks live PvP wire-truth (full battle, multiple bid regressions)
Six distinct fixes accumulated over live-test iterations against four bids
(654473755566, 806245601092, 283192092460, 131549100204, 799755786270) — together
they take the shadow engine from "throws on the first non-mulligan play" to
"survives a full PvP battle, only weird-edge-case Unity touches still left to whack".

1. Engine StableRandom seed aligned with clients' Matched.seed
   (BattleSession.EnsureEngineSetup, NodeNativeBattleHarness.Create). Clients seed
   _stableRandom with BattleSeeds.Stable(masterSeed) (the value the node ships in
   Matched.seed); we were passing the RAW masterSeed to engine.Setup, so every
   StableRandom call diverged from call #1 onward — every turn-1+ draw picked a
   different deck position than the clients. Verified Stable(1184631275)=1543475792
   matches the wire on bid 654473755566.

2. SeedDeck advances cardTotalNum to deck.Count+1 + pins BattleStartDeckCardList.
   Mirrors SBattleLoad.InitPlayer's tail (SBattleLoad.cs:1292). Without it,
   skill-generated tokens auto-assigned Index 0,1,... and COLLIDED with deck-loaded
   indices 1..40 — silent until something addressed the deck card with the
   colliding Index (Hoverboarder at deck idx 1 + a token at engine Index 1 made
   GetBattleCardIdx's SingleOrDefault throw on bid 806245601092).

3. BattleCardView.GameObject lazily non-null in the shim (ViewUiTouchStubs.cs).
   The IsRecovery card-create delegate (NetworkBattleManagerBase.cs:379) passes
   null cardGameObject; Skill_metamorphose.cs:147 in the in-play branch then NRE'd
   on `metamorphosedCard.BattleCardView.GameObject.transform.rotation = identity`,
   a purely cosmetic touch with no game-state implication. Bid 283192092460:
   Petrification on a board follower.

4. TranslateChoiceKeyAction unwraps wrapped selectCard on shadow ingest
   (SessionBattleEngine.cs, sibling to TranslateTargetOwners). Live sender-send
   wires Choice plays as selectCard:{cardId:[...], open:0}; engine's
   ConvertToListInt does `value as List<object>` — a Dict casts to null and
   foreach NREs. The receiver's swallow-all catch (NetworkBattleReceiver.cs:1255)
   logs to Debug.LogError + LocalLog — both shimmed/no-op'd headlessly — and
   returns false, but Receive calls ReceivedMessage with checkBreakData:false so
   the false isn't propagated. The play continues with choiceIdList=[], the chosen
   branch never resolves, the played card stays in hand; a later targeted play
   (A's bounce on B's "board" idx 20) then can't find the target → NRE on null in
   ActionProcessor.PlayCard:407. Bid 131549100204: B's Resonance + A's bounce.
   Opponent-relay path is unaffected — node strips selectCard from broadcasts.

5. HeadlessHandViewStub overrides HandUnfocus/HandFocus/FocusRearrangeHandHand
   to return NullVfx. CreateHandControl returns null in headless; the base
   methods unconditionally deref `_handControl.SetHandState(...)`. A follower
   with a when_spell_play Heal trigger fired on its leader for amount 0 — even
   a 0-heal drives ApplyHealing → CreatePullHandInVfx → HandUnfocus → NRE.
   Bid 799755786270: two consecutive spell plays both crashed this stack.
   Added InternalsVisibleTo("SVSim.BattleEngine.Tests") so the shim-level
   regression tests can pin the no-op contracts directly.

Plus the previous-session fixes carried in this same uncommitted state
(see docs/superpowers/plans/2026-06-07-shadow-engine-desync-handoff.md):
  - doesPlayerGoFirst:true + mgr.IsFirst:true (turn-1 draw count correct
    per seat)
  - RecoveryOperationCollection.PlayHandCardOperation routes all type:30
    through PlaySkillSelectHandCardOperation (skips the two-phase user-select
    guard that aborts targeted spells in recovery)
  - ShadowFeed + ToRawBody: server-generated typed bodies (DealBody, etc.)
    converted to RawBody before engine.Receive (`env.Body as RawBody`
    returned null for typed bodies)
  - Ready idxChangeSeed seeds A's XorShift via the receiver; B's seed is
    injected via SeedOppoIdxChange (BattleSeeds.IdxChange + viewerId)
  - ReadySpin defaulted to 0 (was 243) — non-zero double-cranks the shadow
    which ingests BOTH sides' Ready frames on one stream

Test counts: SVSim.UnitTests 1054/1054, SVSim.BattleEngine.Tests 34/34.

Open: known-residual Unity touches are individual whack-a-mole now (per-card
skill edge cases), not the structural divergences fixed here.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-06-07 19:05:07 -04:00

183 lines
11 KiB
C#

using System.Collections.Generic;
using System.Linq;
using System.Text.Json.Nodes;
using NUnit.Framework;
using SVSim.BattleNode.Protocol;
using SVSim.BattleNode.Sessions.Engine;
namespace SVSim.BattleEngine.Tests.SessionEngine
{
/// <summary>
/// PHASE 4 STEP 1 — Tier 2 capture-replay root-cause VERIFICATION (NOT a fix).
///
/// Replays the FRESH smoke captures (battle 907324319325) — battle_test_fresh_cl1/cl2.ndjson —
/// through a <see cref="SessionBattleEngine"/>, then measures whether the per-seat <c>idxChangeSeed</c>
/// the real Ready frame carries is what controls the "Target card was not found in hand cards"
/// divergence symptom.
///
/// FAITHFUL SETUP (the live ShadowIngest only feeds client SENDS, which contain NO Deal/Ready, so a
/// bare send-only replay can't even seat a hand — that conflates "missing Deal" with "missing
/// reshuffle"). To ISOLATE the reshuffle/seed effect we seat each seat's hand from its OWN client's
/// RECEIVE Deal + Swap + Ready (the frames that establish the hand and reach mulligan-end), then replay
/// both clients' interleaved SENDS (the plays). The Ready frame natively carries the real per-seat
/// idxChangeSeed (cl1=1430655717, cl2=661650374), and the engine's recovery receiver calls
/// <c>CreateXorShift</c> from it (NetworkBattleReceiver.cs:1125-1126). The A/B is then:
/// • WITH-SEED: ingest the Ready frame verbatim (idxChangeSeed present) -> XorShift active;
/// • SEED-STRIPPED: ingest the SAME Ready frame with idxChangeSeed forced to -1 -> XorShift inactive
/// (this is exactly the live shadow's effective state, since it never ingests the seed-bearing Ready).
/// The ONLY difference between the two runs is whether the seed reaches CreateXorShift.
///
/// DECK SETUP MECHANISM (feasibility crux, RESOLVED): each side's deck is reconstructed from the
/// capture's <c>Matched.selfDeck</c> (idx-&gt;cardId, the exact shuffled order the node also handed the
/// client) via <see cref="CaptureReplay.SelfDeckFrom"/>; the master seed from <c>Matched.selfInfo.seed</c>.
/// The deck IS in the socket capture — no external fixture needed.
/// </summary>
[TestFixture]
[NonParallelizable]
public class CaptureReplayReshuffleRootCauseTests
{
private static readonly HashSet<string> SkipUris = new()
{
nameof(NetworkBattleUri.Echo),
nameof(NetworkBattleUri.ChatStamp),
nameof(NetworkBattleUri.Gungnir),
};
private static readonly HashSet<string> MulliganUris = new()
{
nameof(NetworkBattleUri.Deal),
nameof(NetworkBattleUri.Swap),
nameof(NetworkBattleUri.Ready),
};
private sealed record ReplayOutcome(
int FrameCount, List<string> Divergences, bool AllDivergencesPostMulligan, bool SelfXorShiftActive);
// Re-parse a captured frame, overriding the Ready body's idxChangeSeed (and oppoIdxChangeSeed if
// present). Used to STRIP the seed (-1) to model the live shadow's seed-less state.
private static MsgEnvelope OverrideReadySeed(CapturedFrame f, int newSeed)
{
var obj = JsonNode.Parse(f.RawBody)!.AsObject();
obj["idxChangeSeed"] = newSeed;
if (obj.ContainsKey("oppoIdxChangeSeed")) obj["oppoIdxChangeSeed"] = newSeed;
return MsgEnvelope.FromJson(obj.ToJsonString());
}
/// <summary>Seat both hands from each client's receive Deal+Swap+Ready, then replay both clients'
/// interleaved SENDS. <paramref name="stripSeed"/> forces the Ready idxChangeSeed to -1 (the live
/// shadow's effective state). Returns divergences + the post-setup self XorShift state.</summary>
private static ReplayOutcome Replay(bool stripSeed)
{
var cl1 = CaptureReplay.Load("battle_test_fresh_cl1.ndjson");
var cl2 = CaptureReplay.Load("battle_test_fresh_cl2.ndjson");
var deckA = CaptureReplay.SelfDeckFrom(cl1);
var deckB = CaptureReplay.SelfDeckFrom(cl2);
Assert.That(deckA, Is.Not.Empty, "cl1 Matched.selfDeck must reconstruct seat A's deck");
Assert.That(deckB, Is.Not.Empty, "cl2 Matched.selfDeck must reconstruct seat B's deck");
var engine = new SessionBattleEngine();
engine.Setup(masterSeed: CaptureReplay.SeedFrom(cl1), seatADeck: deckA, seatBDeck: deckB);
Assert.That(engine.IsReady, Is.True, "engine must seat from the captured decks + seed");
var divergences = new List<string>();
bool sawMulliganEnd = false;
bool anyDivergencePreMulligan = false;
void Ingest(MsgEnvelope env, bool seat, string uri)
{
if (MulliganUris.Contains(uri)) sawMulliganEnd = true;
var r = engine.Receive(env, isPlayerSeat: seat);
if (r.Diverged)
{
divergences.Add($"seat={(seat ? "A" : "B")} {uri}: {Trim(r.RejectReason)}");
if (!sawMulliganEnd) anyDivergencePreMulligan = true;
}
}
// --- Phase 1: seat both hands from the receive setup frames ----------------------------------
// A single Deal seats BOTH opening hands (cl1's receive Deal carries self=A + oppo=B), so we
// ingest Deal ONCE (as seat A) — ingesting both clients' Deals would double-deal (NRE / "Sequence
// contains more than one"). Each seat's Swap then applies that seat's mulligan, and each seat's
// Ready carries THAT seat's idxChangeSeed (cl1's for A, cl2's for B; the recovery receiver consumes
// only the SELF seed per ingest, NetworkBattleReceiver.cs:1126), reaching mulligan-end per seat.
CapturedFrame Receive(IReadOnlyList<CapturedFrame> frames, string uri) =>
frames.First(f => f.Direction == "receive" && f.Uri == uri);
// Deal once (seat A's receive Deal seats both hands).
Ingest(Receive(cl1, nameof(NetworkBattleUri.Deal)).Env, seat: true, nameof(NetworkBattleUri.Deal));
// Each seat's mulligan swap, then each seat's Ready (its own seed).
foreach (var (frames, seat) in new[] { (cl1, true), (cl2, false) })
{
Ingest(Receive(frames, nameof(NetworkBattleUri.Swap)).Env, seat, nameof(NetworkBattleUri.Swap));
var ready = Receive(frames, nameof(NetworkBattleUri.Ready));
var readyEnv = stripSeed ? OverrideReadySeed(ready, -1) : ready.Env;
Ingest(readyEnv, seat, nameof(NetworkBattleUri.Ready));
}
bool selfActive = engine.SelfXorShiftActive;
// --- Phase 2: replay both clients' interleaved SENDS (the plays / turn ops) -------------------
var sends = CaptureReplay.InterleavedSends(cl1, cl2)
.Where(x => !SkipUris.Contains(x.Env.Uri.ToString()))
.ToList();
foreach (var (env, seat) in sends)
Ingest(env, seat, env.Uri.ToString());
return new ReplayOutcome(
FrameCount: sends.Count, divergences, !anyDivergencePreMulligan, selfActive);
}
private static string Trim(string? s) =>
(s ?? "").Split(" @ ")[0];
[Test]
public void Capture_replay_reproduces_post_mulligan_divergence_and_pins_what_the_seed_does_not_fix()
{
var withSeed = Replay(stripSeed: false);
var stripped = Replay(stripSeed: true);
TestContext.WriteLine($"WITH-SEED (Ready idxChangeSeed present): selfXorShiftActive={withSeed.SelfXorShiftActive} " +
$"playFrames={withSeed.FrameCount} divergences={withSeed.Divergences.Count}");
foreach (var d in withSeed.Divergences) TestContext.WriteLine(" DIVERGE " + d);
TestContext.WriteLine($"SEED-STRIPPED (idxChangeSeed=-1, the live shadow state): selfXorShiftActive={stripped.SelfXorShiftActive} " +
$"playFrames={stripped.FrameCount} divergences={stripped.Divergences.Count}");
foreach (var d in stripped.Divergences) TestContext.WriteLine(" DIVERGE " + d);
Assert.Multiple(() =>
{
// (1) The reported symptom reproduces DETERMINISTICALLY from the captures: the replay diverges,
// including the verbatim "Target card was not found in hand cards" exception.
Assert.That(withSeed.Divergences, Is.Not.Empty,
"the capture replay must reproduce the divergence symptom");
Assert.That(withSeed.Divergences.Any(d => d.Contains("not found in hand")), Is.True,
"the reported 'Target card was not found in hand cards' symptom must reproduce");
// (2) All divergences occur AFTER the mulligan barrier — consistent with a post-mulligan cause.
Assert.That(withSeed.AllDivergencesPostMulligan, Is.True, "with-seed divergences are post-mulligan");
Assert.That(stripped.AllDivergencesPostMulligan, Is.True, "stripped divergences are post-mulligan");
// (3) The wire seed DOES drive the engine's XorShift gate (NetworkBattleReceiver.cs:1126):
// present -> active, stripped (the live shadow's state) -> inactive.
Assert.That(withSeed.SelfXorShiftActive, Is.True,
"ingesting the real Ready (idxChangeSeed present) activates the engine's XorShift");
Assert.That(stripped.SelfXorShiftActive, Is.False,
"stripping idxChangeSeed (the live shadow's state) leaves the XorShift inactive");
// (4) THE KEY VERIFICATION FINDING — activating the XorShift via the wire seed does NOT, on its
// own, change the divergence count. The engine's recovery/watch RECEIVE path never performs
// the post-mulligan full-deck reshuffle the live client does: the XorShift's GetChangeInt is
// consumed ONLY by AddToDeckCardIndexChange (BattlePlayerBase.cs:3079) for cards added to the
// deck AFTER mulligan-end, and the per-turn draw is engine-computed off the (un-reshuffled)
// deck order, not driven by the wire's `move idx`. So "feed the seed" alone does NOT fix the
// desync headless — the eventual fix must also make the engine reshuffle the deck post-
// mulligan to match the client (or drive the draw from the wire idx). We PIN this here.
Assert.That(stripped.Divergences.Count, Is.EqualTo(withSeed.Divergences.Count),
"VERIFIED: activating the XorShift via the wire seed alone does NOT change the divergence " +
"count — the engine's receive path does not reshuffle the deck, so the seed is necessary " +
"but NOT sufficient (the fix needs the reshuffle too, not just the seed)");
});
}
}
}