fix(engine): voice 句尾剪字真根因根治(三修:源尾垫静音冲刷 SoundTouch 未冲刷小节)+ onEnd 完成信号 + live 探针适配
- 真根因(live span 8.57s < 文件 8.80s 铁证):SoundTouch Stretch 源结束时不冲刷 内部最后一个小节(~1 sequence),句尾 0.1-0.25s 从未交付音频图——任何平台皆然, ASR 因容错转写仍 sim=1(转写评审盲区的又一实证)。一修固定 300ms 排空窗只是 把这段不存在的交付空等成静音;二修 onEnd 保证「已交付必播完」但交付不了这段 - 三修:源尾垫 VOICE_TAIL_PAD_SEC(0.25s) 静音把真实句尾推出管线;完成 = onEnd (平台无关,disconnect 已交付帧仍播);1.5s 兜底;顶掉复核移至新句数据访问前 (垫长拷贝的 getChannelData 先于杀旧句会打乱探针时序) - live 探针适配:尾部垫使 SoundTouch 读未注册 buffer、gcd 抽读信号消失—— 清扫器加「仍有连接中语音 SPN 则不关段」守卫,断开即移出追踪集合 - engine-voice.test:PAD+I1/I2/I2b/I3 五组 19 断言(含垫长存在性、onEnd 结算、 兜底不早于 1.5s);10 套件全绿 - live 复验(男声 p1):span 8.57→9.04s(≥文件 8.80s),ASR sim=1 verdict=OK
This commit is contained in:
@@ -1,20 +1,25 @@
|
||||
#!/usr/bin/env node
|
||||
/**
|
||||
* engine-voice.test.mjs — 引擎 voice 句尾保真回归(2026-09 真机人耳反馈固化)。
|
||||
* engine-voice.test.mjs — 引擎 voice 句尾保真回归(2026-09 真机人耳反馈三修固化)。
|
||||
*
|
||||
* 固化的教训:voice 自然结束**不能**用墙钟定时器判定(SPN 4096帧@48k≈85ms +
|
||||
* SoundTouch 固有延迟 ≈190ms > 任何小余量),必须以「源真实消耗位置」为准,
|
||||
* 耗尽后留排空窗再 disconnect。本测试用可手动驱动的 ScriptProcessorNode mock
|
||||
* 模拟真实消耗,把该语义钉死,防止未来重构回潮。
|
||||
* 固化的教训(三轮排查递进):
|
||||
* 一修(废止):源消耗位置 + 固定 300ms 排空窗。手机端未交付量随平台回调
|
||||
* 相位浮动(~0.5s),末词仍被剪。
|
||||
* 二修(保留):完成信号改用 SoundTouch onEnd(extract 返回 0 = 管线内已无
|
||||
* 未交付帧),平台无关;定时器仅剩「源未耗尽等待」与「onEnd 缺失兜底」。
|
||||
* 三修(根治):live 实测发现 SoundTouch Stretch 源结束时不冲刷内部最后
|
||||
* 一个小节——句尾 0.1-0.25s **根本不会被交付**(任何平台,ASR 因容错听不出)。
|
||||
* 根治 = 源尾垫 VOICE_TAIL_PAD_SEC(0.25s) 静音把真实句尾推出管线。
|
||||
*
|
||||
* 三条不变式:
|
||||
* I1 源未耗尽时,_finishVoice 提前触发不得结算(必须短重排等待);
|
||||
* I2 源耗尽后第一次触发只挂排空旗,排空窗(VOICE_DRAIN_MS)过后才真结算;
|
||||
* I3 消耗位置恒为 0(playtest mock / SPN 首回调未到)走旧立即结算路径。
|
||||
* 四条不变式:
|
||||
* I1 源未耗尽时,_finishVoice 提前触发不得结算(短重排等待);
|
||||
* I2 源(含尾部垫)耗尽后须等 onEnd(管线真排空)才结算;onEnd 一到立即(≤~120ms)结算;
|
||||
* I2b onEnd 异常缺失时,兜底定时器在 VOICE_DRAIN_FALLBACK_MS 后结算(宁迟勿剪);
|
||||
* I3 消耗位置恒 0(playtest mock / SPN 首回调未到)走旧立即结算路径。
|
||||
* 附加 PAD:播放源长度 = 原时长 + 0.25s(尾部静音垫存在性)。
|
||||
*
|
||||
* 运行:tools/node/bin/node web/harness/engine-voice.test.mjs
|
||||
*/
|
||||
import fs from 'node:fs';
|
||||
import path from 'node:path';
|
||||
import { fileURLToPath } from 'node:url';
|
||||
|
||||
@@ -26,29 +31,29 @@ const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
|
||||
|
||||
// ---- 全局 mock(在 import 引擎前就位)----
|
||||
const storageMap = new Map();
|
||||
globalThis.window = {
|
||||
AudioContext: null, // 下面赋值
|
||||
addEventListener() {},
|
||||
};
|
||||
globalThis.window = { AudioContext: null, addEventListener() {} };
|
||||
globalThis.localStorage = {
|
||||
getItem: (k) => (storageMap.has(k) ? storageMap.get(k) : null),
|
||||
setItem: (k, v) => storageMap.set(k, String(v)),
|
||||
removeItem: (k) => storageMap.delete(k),
|
||||
};
|
||||
globalThis.document = { addEventListener() {}, getElementById: () => null, body: { appendChild() {}, classList: { add() {}, remove() {} } } };
|
||||
try { globalThis.navigator = { userAgent: 'test', vibrate: () => true }; } catch (_) { /* Node22 自带只读 navigator */ }
|
||||
try { globalThis.navigator = { userAgent: 'test', vibrate: () => true }; } catch (_) { /* Node22 只读 navigator */ }
|
||||
globalThis.performance = globalThis.performance || { now: () => Date.now() };
|
||||
|
||||
const SR = 44100;
|
||||
const DUR = 2.0; // 测试语音 2 秒
|
||||
const DUR = 2.0; // 解码出的原始时长
|
||||
const PAD = 0.25; // 引擎尾部静音垫(VOICE_TAIL_PAD_SEC)
|
||||
const SRC_DUR = DUR + PAD; // 播放源总时长
|
||||
const FRAMES = Math.round(SR * DUR);
|
||||
const SRC_FRAMES = Math.round(SR * SRC_DUR);
|
||||
|
||||
class FakeParam { constructor(v = 1) { this.value = v; } setValueAtTime() {} linearRampToValueAtTime() {} cancelScheduledValues() {} }
|
||||
|
||||
class DrivableSPN {
|
||||
constructor() { this.connect = () => {}; this.disconnectCount = 0; this.onaudioprocess = null; }
|
||||
disconnect() { this.disconnectCount++; }
|
||||
/** 模拟一次真实音频回调:SoundTouch 从源拉取至多 4096 帧 */
|
||||
/** 模拟一次真实音频回调:SoundTouch 从源拉取至多 4096 帧交付音频图 */
|
||||
pump() {
|
||||
if (!this.onaudioprocess) return 0;
|
||||
const out = { outputBuffer: { getChannelData: () => new Float32Array(4096 * 2) } };
|
||||
@@ -63,74 +68,99 @@ class MockAudioContext {
|
||||
createGain() { return { gain: new FakeParam(), connect() {}, disconnect() {} }; }
|
||||
createDynamicsCompressor() { const n = this.createGain(); n.threshold = new FakeParam(); n.ratio = new FakeParam(); n.attack = new FakeParam(); n.release = new FakeParam(); return n; }
|
||||
createScriptProcessor() { return new DrivableSPN(); }
|
||||
/** 引擎三修:源尾垫静音需要 createBuffer + copyToChannel */
|
||||
createBuffer(channels, length, sampleRate) {
|
||||
const chans = Array.from({ length: channels }, () => new Float32Array(length));
|
||||
return {
|
||||
numberOfChannels: channels, length, sampleRate, duration: length / sampleRate,
|
||||
getChannelData: (i) => chans[i],
|
||||
copyToChannel(src, ch) { chans[ch].set(src); },
|
||||
};
|
||||
}
|
||||
decodeAudioData() {
|
||||
return Promise.resolve({
|
||||
duration: DUR, sampleRate: SR, numberOfChannels: 1,
|
||||
getChannelData: (i) => new Float32Array(FRAMES), // 真实长度:SoundTouch extract 按帧索引
|
||||
duration: DUR, length: FRAMES, sampleRate: SR, numberOfChannels: 1,
|
||||
getChannelData: () => new Float32Array(FRAMES),
|
||||
});
|
||||
}
|
||||
resume() { this.state = 'running'; return Promise.resolve(); }
|
||||
}
|
||||
globalThis.window.AudioContext = MockAudioContext;
|
||||
|
||||
globalThis.fetch = async () => ({ ok: true, status: 200, arrayBuffer: async () => new ArrayBuffer(8) });
|
||||
|
||||
// ---- 导入真实引擎 ----
|
||||
const { WebAudioEngine } = await import(path.join(WEB, 'engine', 'audio.js'));
|
||||
const FALLBACK_MS = 1500;
|
||||
|
||||
const engine = new WebAudioEngine('audio/');
|
||||
await engine.unlock();
|
||||
const completed = [];
|
||||
const makeOnComplete = (tag) => () => completed.push(tag);
|
||||
|
||||
console.log('\n=== PAD:播放源存在尾部静音垫 ===');
|
||||
engine.play('voice/tail_pad', { onComplete: () => completed.push('P') });
|
||||
await sleep(30);
|
||||
let token = engine._voice;
|
||||
assert(token && token.srcBuf && Math.abs(token.srcBuf.duration - SRC_DUR) < 0.01,
|
||||
`播放源 = 原时长+0.25s 垫(${token.srcBuf.duration.toFixed(2)}s = ${DUR}+${PAD})`);
|
||||
engine.stopVoice(); engine._voice = null; completed.length = 0;
|
||||
|
||||
console.log('\n=== I1 源未耗尽:提前触发不结算 ===');
|
||||
engine.play('voice/tail_test', { onComplete: makeOnComplete('A') });
|
||||
await sleep(30); // 等 _playVoice 的 await _load 落地
|
||||
let token = engine._voice;
|
||||
engine.play('voice/tail_test', { onComplete: () => completed.push('A') });
|
||||
await sleep(30);
|
||||
token = engine._voice;
|
||||
assert(token && token.buf && token.filter, 'voice token 已建立(含 filter 引用)');
|
||||
for (let i = 0; i < 3; i++) token.node.pump(); // 消耗 ~12288 帧 ≈ 0.28s << 2s
|
||||
const consumedEarly = token.filter.sourcePosition / token.buf.sampleRate;
|
||||
assert(consumedEarly > 0 && consumedEarly < DUR - 0.02, `消耗位置已推进且未耗尽(${consumedEarly.toFixed(2)}s / ${DUR}s)`);
|
||||
engine._finishVoice(token); // 模拟旧式墙钟定时器提前到点
|
||||
for (let i = 0; i < 3; i++) token.node.pump();
|
||||
const consumedEarly = token.filter.sourcePosition / token.srcBuf.sampleRate;
|
||||
assert(consumedEarly > 0 && consumedEarly < token.srcBuf.duration - 0.02,
|
||||
`消耗位置已推进且未耗尽(${consumedEarly.toFixed(2)}s / ${SRC_DUR}s)`);
|
||||
engine._finishVoice(token);
|
||||
assert(engine._voice === token && !token.stopped, 'I1 未耗尽时不结算(token 仍存活,等待重排)');
|
||||
assert(completed.length === 0, 'I1 onComplete 未被提前调用');
|
||||
|
||||
console.log('\n=== I2 源耗尽:排空窗过后才真结算 ===');
|
||||
console.log('\n=== I2 源(含垫)耗尽:onEnd(管线真排空)才结算 ===');
|
||||
let guard = 0;
|
||||
while (token.filter.sourcePosition / SR < DUR && guard++ < 200) token.node.pump();
|
||||
const consumedFull = token.filter.sourcePosition / token.buf.sampleRate;
|
||||
assert(consumedFull >= DUR - 0.02, `源已耗尽(${consumedFull.toFixed(2)}s)`);
|
||||
if (token.endTimer !== null) { clearTimeout(token.endTimer); token.endTimer = null; } // 摘掉墙钟定时器,纯手动驱动
|
||||
const t0 = Date.now();
|
||||
engine._finishVoice(token); // 第一次:应只挂排空旗
|
||||
assert(token.drained === true && engine._voice === token, 'I2 耗尽后第一触发只置排空旗(不结算)');
|
||||
await sleep(VOICE_DRAIN_PLUS()); // 排空窗
|
||||
engine._finishVoice(token); // 第二次:真结算
|
||||
assert(engine._voice === null && token.stopped, 'I2 排空窗后真结算(_voice 清空)');
|
||||
while (token.filter.sourcePosition < SRC_FRAMES && guard++ < 400) token.node.pump();
|
||||
token.node.pump(); // SPN 源耗尽后继续回调 → vendor extract 返回 0 → onEnd
|
||||
assert(token.streamEnded === true, 'I2 源耗尽后 SPN 继续回调触发 onEnd(filter.callback 已接线)');
|
||||
await sleep(120);
|
||||
assert(engine._voice === null && token.stopped, 'I2 onEnd 后短时间内真结算(≤120ms)');
|
||||
assert(completed.includes('A'), 'I2 onComplete 恰好回调一次');
|
||||
assert(token.node.disconnectCount >= 1, 'I2 结算时节点已 disconnect');
|
||||
assert(Date.now() - t0 >= 280, `I2 排空窗确实等待了 ≥280ms(实测 ${Date.now() - t0}ms)`);
|
||||
function VOICE_DRAIN_PLUS() { return 380; }
|
||||
|
||||
console.log('\n=== I2b 无 onEnd 的异常平台:兜底定时器结算(不剪) ===');
|
||||
engine.play('voice/tail_nofb', { onComplete: () => completed.push('B') });
|
||||
await sleep(30);
|
||||
const t2 = engine._voice;
|
||||
guard = 0;
|
||||
while (t2.filter.sourcePosition < SRC_FRAMES && guard++ < 400) t2.node.pump(); // 耗尽但不再多 pump(无 onEnd)
|
||||
if (t2.endTimer !== null) { clearTimeout(t2.endTimer); t2.endTimer = null; }
|
||||
const fb0 = Date.now();
|
||||
engine._finishVoice(t2); // 墙钟路径:drained + 兜底
|
||||
assert(t2.drained === true && engine._voice === t2, 'I2b 耗尽后墙钟触发只挂兜底(不结算)');
|
||||
await sleep(400);
|
||||
assert(engine._voice === t2, 'I2b 兜底期内(400ms < 1500ms)仍不结算');
|
||||
await sleep(FALLBACK_MS - 400 + 150);
|
||||
assert(engine._voice === null && t2.stopped, `I2b 兜底触发后结算(实测 ${Date.now() - fb0}ms ≥ ${FALLBACK_MS})`);
|
||||
assert(completed.includes('B'), 'I2b onComplete 经兜底回调');
|
||||
|
||||
console.log('\n=== I3 消耗位置恒 0(mock/SPN 未回调):旧立即结算路径 ===');
|
||||
engine.play('voice/tail_zero', { onComplete: makeOnComplete('B') });
|
||||
await sleep(30);
|
||||
const token2 = engine._voice;
|
||||
assert(token2 && token2.filter.sourceSound.position === 0, '第二句从未被 pump(position=0)');
|
||||
engine._finishVoice(token2);
|
||||
assert(engine._voice === null && token2.stopped, 'I3 position=0 时立即结算(playtest mock 兼容,链推进不受影响)');
|
||||
assert(completed.includes('B'), 'I3 onComplete 立即回调');
|
||||
|
||||
console.log('\n=== 附加:打断语义不受影响(新句顶掉旧句,旧句 onComplete 不回调) ===');
|
||||
engine.play('voice/tail_test2', { onComplete: makeOnComplete('C1') });
|
||||
engine.play('voice/tail_zero', { onComplete: () => completed.push('C') });
|
||||
await sleep(30);
|
||||
const t3 = engine._voice;
|
||||
for (let i = 0; i < 3; i++) t3.node.pump();
|
||||
engine.play('voice/tail_test3', { onComplete: makeOnComplete('C2') });
|
||||
assert(t3 && t3.filter.sourcePosition === 0, '该句从未被 pump(position=0)');
|
||||
engine._finishVoice(t3);
|
||||
assert(engine._voice === null && t3.stopped, 'I3 position=0 时立即结算(playtest mock 兼容)');
|
||||
assert(completed.includes('C'), 'I3 onComplete 立即回调');
|
||||
|
||||
console.log('\n=== 附加:打断语义不受影响(新句顶掉旧句,旧句 onComplete 不回调) ===');
|
||||
engine.play('voice/tail_test2', { onComplete: () => completed.push('D1') });
|
||||
await sleep(30);
|
||||
assert(engine._voice && engine._voice !== t3 && t3.stopped, '新 voice 顶掉旧 voice(旧 token 已停)');
|
||||
assert(!completed.includes('C1'), '被打断的旧句不回调 onComplete(排队链语义保持)');
|
||||
engine._finishVoice(engine._voice); if (engine._voice) { await sleep(380); engine._finishVoice(engine._voice); }
|
||||
const t4 = engine._voice;
|
||||
for (let i = 0; i < 3; i++) t4.node.pump();
|
||||
engine.play('voice/tail_test3');
|
||||
await sleep(30);
|
||||
assert(engine._voice && engine._voice !== t4 && t4.stopped, '新 voice 顶掉旧 voice(旧 token 已停)');
|
||||
assert(!completed.includes('D1'), '被打断的旧句不回调 onComplete(排队链语义保持)');
|
||||
assert(t4.streamEnded === false, '被打断句不触发 onEnd 结算路径');
|
||||
engine.stopVoice();
|
||||
engine._voice = null;
|
||||
|
||||
|
||||
@@ -141,8 +141,8 @@
|
||||
};
|
||||
|
||||
const AC = window.AudioContext || window.webkitAudioContext;
|
||||
const voiceNodes = new Set(); // 连接中的 SoundTouch ScriptProcessorNode(断开即移除)
|
||||
if (AC) {
|
||||
const voiceNodes = new WeakSet(); // SoundTouch 管线创建的 ScriptProcessorNode
|
||||
const origCSP = AC.prototype.createScriptProcessor;
|
||||
AC.prototype.createScriptProcessor = function (...a) {
|
||||
const n = origCSP.apply(this, a);
|
||||
@@ -153,12 +153,20 @@
|
||||
AudioNode.prototype.disconnect = function (...a) {
|
||||
// voice 节点断开 = 引擎 killVoice(自然结束/被顶掉/stopVoice 都走这里)。
|
||||
// 玩家实际听到的尾音就到此刻(FIFO 里未播部分被丢弃),用 now() 而非 lastAct。
|
||||
if (voiceNodes.has(this) && seg) closeSeg(now(), 'kill');
|
||||
if (voiceNodes.has(this)) {
|
||||
voiceNodes.delete(this);
|
||||
if (seg) closeSeg(now(), 'kill');
|
||||
}
|
||||
return origDisc.apply(this, a);
|
||||
};
|
||||
}
|
||||
// 兜底清扫:长静默后关闭未结段
|
||||
setInterval(() => { if (seg && now() - seg.lastAct > VOICE_GAP_MS) closeSeg(seg.lastAct, 'end'); }, 200);
|
||||
// 兜底清扫:长静默后关闭未结段。
|
||||
// 守卫(2026-09 三修适配):引擎的尾部静音垫让 SoundTouch 读的是新建
|
||||
// buffer(未注册 url),gcd 抽读信号消失、lastAct 停在句首——只要仍有
|
||||
// 连接中的语音 SPN 就说明句在播,不许清扫;段关闭只信 disconnect。
|
||||
setInterval(() => {
|
||||
if (seg && now() - seg.lastAct > VOICE_GAP_MS && voiceNodes.size === 0) closeSeg(seg.lastAct, 'end');
|
||||
}, 200);
|
||||
} catch (e) { push({ ev: 'probe_error', where: 'voice-detect', msg: String(e) }); }
|
||||
|
||||
// ---------- AudioBufferSourceNode:bgm/sfx 播放事件 ----------
|
||||
|
||||
Reference in New Issue
Block a user