Feat/plugin support, allow third_party feat(spine) to create plugin for creator.

[Tianze-Chen]

Re: #

Changelog

---

Continuous Integration

This pull request:

• needs automatic test cases check.

  Manual trigger with @cocos-robot run test cases afterward.

• does not change any runtime related code or build configuration

  If any reviewer thinks the CI checks are needed, please uncheck this option, then close and reopen the issue.

---

Compatibility Check

This pull request:

• changes public API, and have ensured backward compatibility with deprecated features.
• affects platform compatibility, e.g. system version, browser version, platform sdk version, platform toolchain, language version, hardware compatibility etc.
• affects file structure of the build package or build configuration which requires user project upgrade.
• introduces breaking changes, please list all changes, affected features and the scope of violation.

[Tianze-Chen]

@cocos-robot run test cases

[github-actions]

Code Size Check Report

Wechat (WASM)	Before	After	Diff
2D Empty (legacy pipeline)	1015834 bytes	1021919 bytes	⚠️ +6085 bytes
2D All (legacy pipeline)	2684860 bytes	2690991 bytes	⚠️ +6131 bytes
2D All (new pipeline)	2777515 bytes	2783629 bytes	⚠️ +6114 bytes
(2D + 3D) All	10037760 bytes	10044273 bytes	⚠️ +6513 bytes

Web (WASM + ASMJS)	Before	After	Diff
(2D + 3D) All	16875138 bytes	16881461 bytes	⚠️ +6323 bytes

Interface Check Report

! WARNING this pull request has changed these public interfaces:

@@ -2681,8 +2681,107 @@
         protected _applyFontTexture(): void;
         protected changeMaterialForDefine(): void;
     }
     /**
+     * @en A segment of the mesh: a range of indices drawn with one texture+material.
+     * @zh 网格的一个片段：一段索引，用同一纹理+材质绘制。
+     */
+    export interface UIMeshSegment {
+        indexOffset: number;
+        indexCount: number;
+        texture: Texture2D | null;
+        material: renderer.MaterialInstance | null;
+    }
+    /**
+     * @en Pre-baked mesh data for one frame.
+     * @zh 一帧的预烘焙网格数据。
+     * vertexStride: 24 (single-color V3F_T2F_C4B) or 28 (two-color V3F_T2F_C4B_C4B).
+     */
+    export interface UIMeshData {
+        vertexCount: number;
+        vertexStride: number;
+        vertexData: Uint8Array;
+        indexCount: number;
+        indexData: Uint8Array;
+        segments: UIMeshSegment[];
+    }
+    /**
+     * @en A generic 2D mesh renderer that consumes pre-baked vertex/index data.
+     * The data provider (e.g. a spine plugin) fills setMeshData every frame; this
+     * component handles buffer allocation, batching and submission.
+     * @zh 通用 2D 网格渲染器，消费预烘焙的顶点/索引数据。数据提供方（如 spine 插件）
+     * 每帧调用 setMeshData，本组件负责缓冲分配、合批与提交。
+     */
+    export class UIMesh extends UIRenderer {
+        protected _enableBatch: boolean;
+        /**
+         * @en Whether the incoming color data is premultiplied-alpha. Declares how
+         * cascaded opacity is folded into the vertices (alpha byte only vs
+         * RGBA + dark RGB) and the blend factors of the builtin material.
+         * @zh 输入颜色数据是否为预乘 alpha 格式。决定级联不透明度折算方式
+         * （仅 alpha 字节 vs RGBA + dark RGB）及内置材质混合因子。
+         */
+        protected _premultipliedAlpha: boolean;
+        protected _meshData: UIMeshData | null;
+        protected _useTint: boolean;
+        constructor();
+        /**
+         * @en Feeds the pre-baked mesh data for the current frame.
+         * @zh 喂入当前帧的预烘焙网格数据。
+         */
+        setMeshData(data: UIMeshData): void;
+        /**
+         * @en Whether to enable sprite batching.
+         * @zh 是否启用合批。
+         */
+        get enableBatch(): boolean;
+        set enableBatch(value: boolean);
+        /**
+         * @en Whether the input color data is premultiplied-alpha.
+         * @zh 输入颜色数据是否为预乘 alpha 格式。
+         */
+        get premultipliedAlpha(): boolean;
+        set premultipliedAlpha(value: boolean);
+        /**
+         * Notifies subclasses that the declared data format changed. The property
+         * itself is owned here (single source of truth); data producers — e.g. a
+         * spine plugin whose C++ side premultiplies vertex colors — override this
+         * to forward the format to their baker instead of redeclaring the field.
+         */
+        protected onPremultipliedAlphaChanged(): void;
+        onLoad(): void;
+        protected _updateColor(): void;
+        /**
+         * JSB staleness poll. The native draw-info path re-runs _prepareBuffers
+         * (which bakes the world transform and folds cascaded opacity into the
+         * vertex bytes) only when this renderer is marked dirty — and neither node
+         * transforms nor UIOpacity mark middleware renderers (TRANSFORM_CHANGED
+         * fires only on the changed node, UIOpacity writes localOpacity with no
+         * event). Poll the captured inputs and re-mark on drift. Web runs
+         * _prepareBuffers unconditionally through fillBuffers, so skip there.
+         * Subclasses overriding update should call super.update(dt).
+         */
+        update(dt: number): void;
+        protected _flushAssembler(): void;
+        updateRenderer(): void;
+        protected _render(batcher: any): void;
+        /**
+         * The builtin material is the spine effect (default-spine-material): the
+         * stock ui-sprite-material has no USE_LOCAL variant, so node-local input
+         * through the component material would render stuck at the origin. The
+         * spine effect carries both USE_LOCAL and TWO_COLORED macros; segment
+         * materials supplied via setMeshData are the provider's own contract.
+         */
+        protected _updateBuiltinMaterial(): Material;
+        /**
+         * Keeps the builtin material instance in step with the declared vertex
+         * space / data format: USE_LOCAL for the GPU take-over mode, TWO_COLORED
+         * for two-color data, and blend factors matching the alpha format.
+         */
+        updateMaterial(): void;
+        protected createRenderEntity(): __private._cocos_2d_renderer_render_entity__RenderEntity;
+    }
+    /**
      * @en
      * The Mask Component.
      *
      * @zh
@@ -61397,8 +61496,34 @@
         stop(): void;
     }
     /**
      * @en
+     * The engine's packaged cross-platform WebAssembly interface (pal/wasm).
+     *
+     * Re-exported under the public `cc.wasm` namespace so extension/game code can
+     * load its own `.wasm` files through the same platform-adaptive path the engine
+     * uses internally for box2d / physx / spine / webgpu:
+     *
+     *   - web:      fetch the `.wasm` bytes, then `WebAssembly.instantiate`;
+     *   - mini-game: resolve the path into `cocos-js/` and delegate to the platform's
+     *               `CCWebAssembly.instantiate` (which accepts a file path, never
+     *               raw bytes — this is why embedded-base64 wasm fails there);
+     *   - native:   read the file from `src/cocos-js/` via `fileUtils`.
+     *
+     * The `wasmUrl` argument is a bare file name (e.g. `'foo.wasm'`) whose file is
+     * expected to land in the build output's `cocos-js/` directory.
+     * @zh
+     * 引擎封装好的跨平台 WebAssembly 接口（pal/wasm），通过 `cc.wasm` 命名空间公开，
+     * 供扩展/游戏代码用与引擎内部一致的路径加载自己的 `.wasm`。
+     */
+    export const wasm: {
+        instantiateWasm: typeof __private._pal_wasm__instantiateWasm;
+        fetchBuffer: typeof __private._pal_wasm__fetchBuffer;
+        fetchUrl: typeof __private._pal_wasm__fetchUrl;
+        ensureWasmModuleReady: typeof __private._pal_wasm__ensureWasmModuleReady;
+    };
+    /**
+     * @en
      * WebView component, used to display web pages in the game.
      * Since different platforms have different authorizations, APIs, and control methods for WebView components, there is no unified standard yet.
      * So currently only Web, iOS, and Android platforms are supported.
      * @zh
@@ -63657,8 +63782,12 @@
         export class _cocos_2d_renderer_static_vb_accessor__StaticVBAccessor extends _cocos_2d_renderer_buffer_accessor__BufferAccessor {
             static IB_SCALE: number;
             static ID_COUNT: number;
             get id(): number;
+            /** Per-chunk capacity; allocateChunk refuses requests above these. */
+            get maxVertexCount(): number;
+            /** Per-chunk index capacity; allocateChunk refuses requests above these. */
+            get maxIndexCount(): number;
             constructor(device: gfx.Device, attributes: gfx.Attribute[], vCount?: number, iCount?: number);
             destroy(): void;
             reset(): void;
             getVertexBuffer(bid: number): Float32Array;
@@ -76065,8 +76194,33 @@
             enable(): void;
             disable(noPause?: boolean): void;
             syncMatrix(): void;
         }
+        /**
+         * The first parameter of standard `Webassembly.instantiate` interface is the arraybuffer of wasm.
+         * But the implementation on some platforms is not standard.
+         * So here we provide a more commonly used interface, whose first parameter is the url of wasm.
+         *
+         * @param wasmUrl the url of wasm, this should be a url relative from build output chunk.
+         * @param importObject the standard `WebAssembly.Imports` instance
+         */
+        export function _pal_wasm__instantiateWasm(wasmUrl: string, importObject: WebAssembly.Imports): Promise<WebAssembly.WebAssemblyInstantiatedSource>;
+        /**
+         * Fetch binary data from wasm url or js mem url.
+         * NOTE: This method should only use to instantiate asm.js compiled with `-O2` options,
+         * because not all platforms support instantiate wasm by wasm binary.
+         * eg. WeChat can only instantiate wasm by wasm url.
+         *
+         * @param binaryUrl the url of wasm or js mem, this should be a url relative from build output chunk.
+         */
+        export function _pal_wasm__fetchBuffer(binaryUrl: string): Promise<ArrayBuffer>;
+        export function _pal_wasm__fetchUrl(binaryUrl: string): Promise<string>;
+        /**
+         * Sometimes we need to put wasm modules in subpackage to reduce code size.
+         * In this case we need to ensure that the wasm modules is ready before we import them.
+         * Please remember to invoke this method before we import wasm modules.
+         */
+        export function _pal_wasm__ensureWasmModuleReady(): Promise<void>;
         export enum _cocos_web_view_web_view_enums__WebViewEventType {
             /**
              * @en None.
              * @zh 无。

[github-actions]

@Tianze-Chen, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: FAIL
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/9085ad25-6161-4814-9071-7b3a3c1abaeb/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/32715612479

Task Details

Platform	build	boot	runned	crashScene	FailScene
web-mobile	PASS	PASS	FAIL		graphics-line-join
ios	PASS	PASS	PASS
mac	PASS	PASS	PASS

[github-actions]

@Tianze-Chen, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: PASS
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/089860d8-b139-44d8-83f5-0f82c30ee3ca/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/32715612479

Task Details

[Tianze-Chen]

@cocos-robot run test cases

[github-actions]

@Tianze-Chen, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: FAIL
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/11edd7a1-712a-43dd-8973-e97607b5e3fa/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/36375486319

Task Details

Platform	build	boot	runned	crashScene	FailScene
web-mobile	PASS	PASS	FAIL		graphics-continuous-filling
ios	PASS	PASS	PASS
mac	PASS	PASS	PASS

[github-actions]

@Tianze-Chen, Please check the result of run test cases:

• Test Platform: PR-Test
• Editor Version: v4.0.0
• Task Result: PASS
• Task URL: https://cctest.cocos.org/#/reportview/PR-TEST/339266d5-361a-42e7-ad67-fac9868e48d3/-1
• GitHub Action: https://github.com/cocos/cocos4/actions/runs/36375486319

Task Details

[bofeng-song]

针对 UIMesh 的四处问题提出行内建议；以下基于代码静态审查，尚未运行外部 Spine 插件验证。

[Tianze-Chen]

@cocos-robot run test cases

[bofeng-song]

复查最新提交：容量预留上限和 Accessor 随 Batcher 重建的修复已落实；仍有三处渲染问题，见行内评论。以下基于渲染链路静态审查及抽取 PR 实际方法的最小复现，其中原生共享缓冲生命周期使用模拟，尚未运行外部 Spine 插件或真机画面验证。

[bofeng-song]

[P1] 原生共享索引缓冲重置后，只更新 dirty 网格会覆盖静止网格的索引。

UIRendererManager 只对 dirty renderer 调用 updateRenderer()，而原生 Batcher2d::uploadBuffers() 每帧会 reset mesh buffer 的索引偏移。假设 A、B 共用同一个 buffer：第一帧 A 的索引在 0、B 的索引在 3；第二帧 A 没有新数据、变换或透明度变化，只有 B 更新，那么这里同步到重置后的偏移 0，B 会覆盖 A 的索引，但 A 保留的 draw info 仍指向 0。A 因而绘制 B 的顶点，可能产生错图或重影，关闭合批也不能避免共享缓冲中的覆盖。

用本 PR 的 _prepareBuffers() 配合模拟原生 reset 的最小复现，A 的 [0,1,2] 在第二帧变成了 B 的 [3,4,5]。建议将每帧索引提交与顶点 dirty 更新分离：每帧为所有参与绘制的网格重建索引及 draw-info 偏移，或改为稳定的独立索引区间。请补充两个 UIMesh 中一个暂停、另一个持续更新的原生跨帧用例。

[bofeng-song]

[P2] 合批路径只应用二维仿射变换，仍与非合批的完整节点矩阵不一致。

这里仅更新 X/Y，保留原始 Z，也没有处理输入 Z 对 X/Y 的贡献。例如局部点 (1,2,3)，节点或父节点沿 Z 平移 10：非合批的 GPU 路径得到 (1,2,13)，当前合批方法仍得到 (1,2,3)。绕 X/Y 轴旋转时也会丢失相应的三维变换；纯 XY 平移/缩放、绕 Z 轴旋转则不受影响。

建议使用完整世界矩阵变换 XYZ，并同步更新 update() 中的变换检测（目前也只检查六个二维矩阵分量），保证暂停动画时修改 Z 位移或 X/Y 旋转仍能刷新。请补充节点及父节点存在这些变换时，合批开关前后结果一致的用例。

[bofeng-song]

[P2] 级联透明度漏乘祖先 UI 渲染组件的 color.a。

这里计算的是自身 color.a × 各级 localOpacity，但 Batcher2D.walk() 会在每一级额外乘入该节点 UI 渲染组件的 color.a / 255。例如父节点有 Sprite，Sprite.color.a = 128，子节点 UIMesh 的顶点 alpha、自身 color.a 均为 255，所有 localOpacity 都为 1：标准 UI 级联结果约为 0.502，而此方法返回 1，UIMesh 仍完全不透明。

父节点的 UIOpacity.opacity = 128 已被本次修复覆盖；遗漏的是父级 Sprite/Label 等渲染组件的颜色 alpha。建议与 Batcher2D 的级联规则保持一致，并确保 JSB 的变化检测也包含该因素；增加修改父级组件 color.a（含动画暂停）的测试。