fix(chargen): Campaign CC CC6b-PRE review fix round — F1-F7 + F11 concession rewrite

F1 (BLOCKING, doc-only) — the idle-by-default rationale rested on an unsound
"uninitialized C++ member defaults to 0" argument (heap operator-new memory
is indeterminate, not zero). Verified and replaced with the real evidence:
gmCGAppearancePage::InitializePage @0x0047FDD0 writes an EXPLICIT
this->m_bZoomedIn = 0; at 0x004802C3, immediately after that same function
points the camera at the zoomed-IN per-heritage eye (0x00480286-0x0048029E).
Fixed in all three places: the register's TS-83 retirement clause,
ChargenPreviewAnimator's class doc, ChargenPreviewZoomController.IsZoomedIn's
doc. Recorded the retail quirk this implies: the character starts framed
close-up while not-zoomed-in, so the first Zoom In click (once mounted)
tweens close-eye->close-eye (visually null) while still freezing the
animation — the port reproduces this faithfully.

F2 — ChargenPreviewZoomController and ChargenPreviewAnimator kept
independent _zoomedIn bools synced only via a nullable animator parameter,
risking desync. Retail's m_bZoomedIn is a single field gating both camera
and animation, so the fix makes the animator the sole state owner:
ChargenPreviewZoomController now takes its ChargenPreviewAnimator as a
required constructor dependency, IsZoomedIn reads straight through to it,
and ZoomIn/ZoomOut no longer take a parameter at all — there is no second
bool left to disagree.

F3 — documented the DoRotation counter-clockwise branch's x87-stack
decompiler artifact (BN renders x87_r7_1 = x87_r6_3 at 0x0047CAEB, which
would store delta-degrees instead of the timestamp for CCW only); the port
already stores "now" in both branches, cited against
feedback_bn_decomp_field_names.md.

F4 — ChargenPreviewAnimator.ApplyIdleFrame now double-buffers two
List<MeshRef> instead of allocating fresh every 30fps tick.

F5 — filed docs/ISSUES.md #402 tracking the RetailAnimationCyclePlayback /
LiveEntityAnimationPresenter duplication as an owned post-CC follow-up,
referenced from the new type's own doc.

F6 — reworded the ChargenPreviewEntityBuilder.TryBuild "byte-identical"
claim to result-identical (TryBuildAnimated now also resolves the idle DID
and loads the idle Animation before the wrapper discards them).

F7 — added the missing clockwise >360 clamp test (readable decomp
polarity, unlike F3's CCW artifact).

ALSO — rewrote the CC6b ledger row's m_alternateSetupID MUST-COVER note per
the reviewer's F11 concession: all five write sites belong to gmBarberUI
(the post-creation barber shop), not gmCGAppearancePage, which has no
option-checkbox-equivalent field at all. Added the enclosing-function
citations and an explicit directive that CC6b-mount must NOT build a
crown/no-flame checkbox on the Appearance page.

Tests: ChargenPreviewRotationControllerTests +1 (10 total),
ChargenPreviewZoomControllerTests +2 and every case rewritten for the
required-animator constructor (9 total). Core.Tests 4786/1 skip (unchanged),
Content.Tests 147/0, App.Tests 5152/6 skips (+3) — zero failures in
isolation, full solution Release build green. Two pre-existing flakes
observed across repeated full-solution runs, neither caused by this round
and neither reproducing standalone: Core.Net.Tests' NakEmissionTests loss
soak, and Content.Tests' DecodedTextureCacheTests concurrency race.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Erik 2026-08-15 19:32:48 +02:00
parent 8dfee1118f
commit 1ba22a01a8
10 changed files with 195 additions and 50 deletions

View file

@ -26,6 +26,17 @@ namespace AcDream.App.Rendering;
/// at frame 0 on every transition INTO the playing state for the same
/// reason: <c>set_sequence_animation</c>'s <c>arg3=1</c> clears the sequence
/// before appending, so every <c>StartAnimation</c> call restarts the clip.
/// The DEFAULT-false claim itself rests on <c>gmCGAppearancePage::InitializePage
/// @ 0x0047FDD0</c>'s explicit <c>this-&gt;m_bZoomedIn = 0;</c> at
/// <c>0x004802C3</c> — written immediately after that same function sets the
/// camera to the zoomed-IN per-heritage eye (<c>0x00480286-0x0048029E</c>),
/// not from the ctor simply never touching the field (heap <c>operator new</c>
/// memory is indeterminate, not zero — that argument doesn't hold on its
/// own). One retail quirk this implies: the character starts framed close-up
/// AND not-zoomed-in at the same time, so the FIRST Zoom In click (once
/// mounted) tweens close-eye→close-eye — visually null — while still
/// freezing the animation; the port reproduces this faithfully rather than
/// treating it as a bug.
/// </para>
///
/// <para>
@ -47,6 +58,15 @@ internal sealed class ChargenPreviewAnimator
private float _currFrame;
private bool _zoomedIn;
// Double-buffered so a 30fps Tick doesn't allocate a fresh List<MeshRef>
// every frame: one buffer is whatever Entity.MeshRefs currently points
// at (potentially still being read by the renderer's own Render() call
// for this frame), the other is safe to Clear()+refill for the NEXT
// tick and only gets published once fully populated.
private readonly List<MeshRef> _meshRefsBufferA = [];
private readonly List<MeshRef> _meshRefsBufferB = [];
private bool _nextBufferIsA = true;
public ChargenPreviewAnimator(ChargenPreviewAnimatedBuild build)
{
_build = build ?? throw new ArgumentNullException(nameof(build));
@ -112,7 +132,9 @@ internal sealed class ChargenPreviewAnimator
{
DatReaderWriter.DBObjs.Animation animation = _build.IdleAnimation!;
IReadOnlyList<ChargenPreviewDrawablePart> parts = _build.DrawableParts;
var meshRefs = new List<MeshRef>(parts.Count);
List<MeshRef> meshRefs = _nextBufferIsA ? _meshRefsBufferA : _meshRefsBufferB;
_nextBufferIsA = !_nextBufferIsA;
meshRefs.Clear();
foreach (ChargenPreviewDrawablePart part in parts)
{
bool resolved = RetailAnimationCyclePlayback.TryInterpolatePart(

View file

@ -144,10 +144,14 @@ internal static class ChargenPreviewEntityBuilder
/// resolved body Setup isn't in the dat source (a corrupted/incomplete
/// install — the same failure shape
/// <see cref="DatLiveEntityProjectionMaterializer"/> treats as "drop this
/// spawn"). Unchanged since CC6a — a thin wrapper over
/// <see cref="TryBuildAnimated"/> that keeps this method's existing
/// callers' behavior byte-identical. New code that wants retail's true
/// default (idle loop playing) should call
/// spawn"). Unchanged since CC6a for its RESULT — a thin wrapper over
/// <see cref="TryBuildAnimated"/> that returns exactly the same
/// <c>WorldEntity</c> (rest-posed) this method's existing callers already
/// expect; ALL 3 of those callers' tests still pass unmodified. Not
/// byte-identical internally any more — <see cref="TryBuildAnimated"/>
/// also resolves the idle DID and loads the idle Animation before this
/// wrapper discards them, extra dat work the pre-CC6b method never did.
/// New code that wants retail's true default (idle loop playing) should call
/// <see cref="TryBuildAnimated"/> and wrap the result in a
/// <see cref="ChargenPreviewAnimator"/> instead.
/// </summary>

View file

@ -84,7 +84,19 @@ internal sealed class ChargenPreviewRotationController
/// <c>[0, 360)</c> — not a full modulo loop; retail's own tail only
/// adds/subtracts 360 once (pseudo-C ~0x0047caf3-0x0047cb31), which is
/// exactly enough for any realistic per-frame delta and is reproduced
/// here verbatim rather than "improved" into a `%=`.
/// here verbatim rather than "improved" into a `%=`. Fix round F3: Binary
/// Ninja literally renders <c>x87_r7_1 = x87_r6_3</c> at <c>0x0047CAEB</c>
/// inside the counter-clockwise branch — reassigning the local that held
/// the "now" timestamp to the just-computed delta-degrees value — which
/// would make the <c>0x0047CB3D</c> store into <c>m_dLastRotateTime</c>
/// write delta-degrees instead of the timestamp for CCW only; that is an
/// x87-FPU-stack modeling artifact of the decompiler, not real retail
/// behavior (a shipped feature where every counter-clockwise rotation
/// visibly diverges from clockwise is implausible, and
/// <c>claude-memory/feedback_bn_decomp_field_names.md</c> names exactly
/// this x87-stack-register mislabeling as a known decompiler artifact
/// class), so this port stores <c>now</c> into <c>_lastRotateTime</c>
/// unconditionally in BOTH directions.
/// </summary>
public void Tick(double now)
{

View file

@ -13,6 +13,24 @@ namespace AcDream.App.Rendering;
/// finishes (see <see cref="ChargenPreviewAnimator"/>'s own doc comment).
///
/// <para>
/// <b>One owner of the zoom state (fix round F2):</b> retail's
/// <c>m_bZoomedIn</c> is a SINGLE field on <c>gmCGAppearancePage</c> that
/// gates both the camera target AND the animation swap — there is no way
/// for retail's own camera and animation to disagree about which zoom state
/// they're in. The first cut of this port kept two independent bools (one
/// here, one on <see cref="ChargenPreviewAnimator"/>) synced only by
/// <see cref="ZoomIn"/>/<see cref="ZoomOut"/> calling a NULLABLE animator
/// parameter — a null pass, or any direct
/// <see cref="ChargenPreviewAnimator.SetZoomedIn"/> call bypassing this
/// controller, would desync the camera's target from the animation's pose.
/// This class now takes its <see cref="ChargenPreviewAnimator"/> as a
/// REQUIRED constructor dependency and <see cref="IsZoomedIn"/> reads
/// straight through to <see cref="ChargenPreviewAnimator.IsZoomedIn"/> — the
/// animator is the sole state owner, matching retail's own single-field
/// design, and there is no longer a second bool that could disagree with it.
/// </para>
///
/// <para>
/// Retail drives <see cref="Tick"/> once per frame from a global-message-3
/// tick while <c>m_bShouldZoomAnimate</c> is set
/// (<c>gmCGAppearancePage::ListenToGlobalMessage @ 0x0047CED0</c>); the
@ -38,59 +56,68 @@ internal sealed class ChargenPreviewZoomController
private const double InvalidDurationSentinel = -0.1;
private readonly uint _heritageId;
private readonly ChargenPreviewAnimator _animator;
private Vector3 _startEye;
private Vector3 _targetEye;
private double _animStartTime;
private double _animDuration;
private bool _shouldAnimate;
private bool _zoomedIn;
public ChargenPreviewZoomController(uint heritageId, ChargenPreviewCamera camera)
public ChargenPreviewZoomController(uint heritageId, ChargenPreviewCamera camera, ChargenPreviewAnimator animator)
{
ArgumentNullException.ThrowIfNull(camera);
ArgumentNullException.ThrowIfNull(animator);
_heritageId = heritageId;
Camera = camera;
_animator = animator;
}
public ChargenPreviewCamera Camera { get; }
/// <summary>Mirrors retail's <c>m_bZoomedIn</c> — false (not zoomed in)
/// is the ctor-implicit default, matching <see cref="ChargenPreviewAnimator"/>'s
/// own default (see that class's doc comment for the shared citation).</summary>
public bool IsZoomedIn => _zoomedIn;
/// <summary>
/// Mirrors retail's <c>m_bZoomedIn</c> — a straight read-through to
/// <see cref="ChargenPreviewAnimator.IsZoomedIn"/> (see this class's own
/// "one owner" doc above), which itself defaults false per
/// <c>gmCGAppearancePage::InitializePage @ 0x0047FDD0</c>'s explicit
/// <c>this-&gt;m_bZoomedIn = 0;</c> at <c>0x004802C3</c> — written right
/// after that same function points the camera at the zoomed-IN
/// per-heritage eye (<c>0x00480286-0x0048029E</c>). One retail quirk
/// this produces: the character starts framed close-up while
/// NOT-zoomed-in, so the first Zoom In click (once mounted) tweens
/// close-eye→close-eye — visually null — while still freezing the
/// animation; this port reproduces it faithfully.
/// </summary>
public bool IsZoomedIn => _animator.IsZoomedIn;
/// <summary>
/// <c>gmCGAppearancePage::ZoomIn @ 0x0047CF00</c>: no-op if already
/// zoomed in (retail's own early-return guard). Otherwise starts a tween
/// from the camera's CURRENT eye to the default (zoomed-IN) per-heritage
/// profile and swaps <paramref name="animator"/> to the frozen rest
/// pose IMMEDIATELY (<c>gmCG3DView::StopAnimation</c>'s call site,
/// pseudo-C ~0x0047d024, precedes the tween's own completion by
/// definition — it runs once, synchronously, inside <c>ZoomIn</c>
/// itself).
/// profile and swaps the animator to the frozen rest pose IMMEDIATELY
/// (<c>gmCG3DView::StopAnimation</c>'s call site, pseudo-C ~0x0047d024,
/// precedes the tween's own completion by definition — it runs once,
/// synchronously, inside <c>ZoomIn</c> itself).
/// </summary>
public void ZoomIn(ChargenPreviewAnimator? animator)
public void ZoomIn()
{
if (_zoomedIn)
if (IsZoomedIn)
return;
StartTween(ChargenPreviewCamera.ResolveDefaultEye(_heritageId));
_zoomedIn = true;
animator?.SetZoomedIn(true);
_animator.SetZoomedIn(true);
}
/// <summary>
/// <c>gmCGAppearancePage::ZoomOut @ 0x0047D050</c>: no-op if not
/// currently zoomed in. Otherwise starts a tween toward the zoomed-OUT
/// per-heritage profile and swaps <paramref name="animator"/> back to
/// the playing idle loop immediately, mirroring <see cref="ZoomIn"/>.
/// per-heritage profile and swaps the animator back to the playing idle
/// loop immediately, mirroring <see cref="ZoomIn"/>.
/// </summary>
public void ZoomOut(ChargenPreviewAnimator? animator)
public void ZoomOut()
{
if (!_zoomedIn)
if (!IsZoomedIn)
return;
StartTween(ChargenPreviewCamera.ResolveZoomedOutEye(_heritageId));
_zoomedIn = false;
animator?.SetZoomedIn(false);
_animator.SetZoomedIn(false);
}
private void StartTween(Vector3 targetEye)

View file

@ -34,7 +34,8 @@ namespace AcDream.Core.Physics;
/// behavior-preserving mechanical follow-up (not done here — that file is
/// live, heavily tested production entity-rendering code with zero relation
/// to this preview-only feature, so touching it is out of this slice's
/// blast radius by design, not oversight).
/// blast radius by design, not oversight). Tracked as
/// <c>docs/ISSUES.md</c> #402 so the follow-up has an owner.
/// </para>
/// </summary>
public static class RetailAnimationCyclePlayback