Controls now tracks whether the player's last deliberate press was a
gamepad button/stick or a keyboard key/mouse click (ignoring plain
mouse motion, since the mouse still drives the camera regardless of
which device moves the bull). The HUD's ability-bar keycaps read from
that: keyboard letters normally, Xbox-style face-button glyphs while
a gamepad is in use, swapping live the moment the player switches
devices.
Godot's joypad button indices are already Xbox-position-normalized
(by SDL natively, by the browser's Standard Gamepad Mapping on web),
so the glyph mapping (button 0/1/2/3 -> A/B/X/Y) reads correctly for
any controller brand with no per-device special-casing.
Verified headlessly: labels start as the bound keyboard letters
(F/G/H/J), switch to X/A/B/Y on a synthetic joypad button press
(matching project.godot's kick/dash/slam/roll bindings), and switch
back on a keyboard press.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXBNaEBmrYDSMcv4NgUWtV
Siim reported controller vibration stops working entirely after the
first playthrough. The gate's pulsed rumble (vibrate_for) fired a new
~0.3s vibration every 0.4s for 7 seconds — about 17 back-to-back start
calls. On at least one setup that appears to leave the gamepad's
rumble motor stuck reporting "busy", so every later vibrate() call
(hits included) silently does nothing for the rest of that session.
Space pulses out more (fewer, longer gaps: 0.25s on / 0.6s off instead
of 0.3s on / 0.1s off) and explicitly stop_joy_vibration() between
them on native instead of relying on each pulse's own duration to
time out, so a new effect never starts before the last one has
cleanly ended.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXBNaEBmrYDSMcv4NgUWtV
Siim reported gamepad menu navigation works but pressing a button
does nothing. Checked Godot's actual built-in ui_accept bindings
directly: only Enter/Kp Enter/Space — no joypad button at all. That's
why D-pad/stick focus movement (ui_left/right/up/down do default to
gamepad input) worked while "confirm" never could, on any controller.
Binds both the Xbox-style bottom face button (A, index 0) and the
right-hand one (B, index 1 — physically labelled "A" on a
Nintendo-layout pad, since gamepad mapping normalizes to Xbox
*position* rather than the printed label), so either common "confirm"
button activates the focused menu/end-screen button.
Verified headlessly end-to-end: a synthetic joypad button 0 or button
1 press on the focused end-screen button now reloads the run, matching
what a real gamepad press should do.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXBNaEBmrYDSMcv4NgUWtV
Siim reported the hit-vibration works but the lobby gate's rumble
doesn't. Godot's start_joy_vibration() docs call this out directly:
long single effects are unreliable on much gamepad hardware/drivers,
and recommend a short effect repeated instead. Added
Controls.vibrate_for() to do that and pointed the gate at it in place
of its single 7-second call.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXBNaEBmrYDSMcv4NgUWtV
- The finish screen's continue button now reads "Play Again" and starts
a fresh run instead of "Continue"-ing to a lobby gate that no longer
exists past the run's last fight.
- The "Your Time" readout moves from the top (overlapping the bull in
thankyou.ogv) to just under the clip's own baked title, restyled to
match its flat dark-ink look instead of the gold HUD outline, and
pulses on the same ~1.4s breathing cycle sampled off the source clip
so it reads as part of the poster rather than a readout stamped on it.
- Added a `video <win|loss|finish>` debug console command to preview
any outcome screen without playing through a fight, via a new
hud.gd debug entry point (`debug_force_outcome`).
- Dropped the console's per-frame "focus stolen by ... — reclaiming"
log line — its own comment flagged it as temporary diagnostic
instrumentation, and it fired on every console open.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXBNaEBmrYDSMcv4NgUWtV
Input.start_joy_vibration() is a documented engine no-op on Godot's
HTML5 export (never wired to the browser side). Controls.vibrate()
wraps it: native platforms go straight through as before, web drives
the browser's Gamepad Haptic Actuator API directly via JavaScriptBridge.
Gamepad movement/ability input already worked on desktop browsers via
the standard Web Gamepad API, so only vibration needed a fix.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXBNaEBmrYDSMcv4NgUWtV
Move every script and scene into role-based directories (actors/, systems/, ui/,
levels/, props/) and all raw art under art/ (grouped by role, snake_case,
Extract-Textures GLBs), dropping Blender version-suffix noise. Filenames are
snake_case (PascalCase reserved for class_name) to avoid case-insensitive
filesystem aliasing on the web/macOS targets. Routes the bull's hits through the
typed DamageEvent pipeline and wires the bull's abilities to the Fx library.
Updates CLAUDE.md's folder-layout overview to match.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CCyaJPTHCH9EDthpf4A5Sj
An Fx autoload + one_shot_effect base so every effect is an inspector-editable
.tscn (the systems/fx/ authoring dir): blood/slam/spark/flash primitives and the
bull's kick/dash/slam/roll + movement effects (bull_*.tscn), fired via
Fx.spawn() / Fx.attach(). Plus a preview gallery + smoke test, and
bull_impact_fx.gd for the coded, non-particle slam rings + kick flash-pop.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CCyaJPTHCH9EDthpf4A5Sj
A typed DamageEvent contract (systems/combat/damage_event.gd) as a single
entry point for every attacker, plus a characterization test locking the bull's
damage reception. Attacker/receiver wiring rides along with the moved actor
scripts in the reorg commit.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CCyaJPTHCH9EDthpf4A5Sj
richard