Address cube, validation, and Meshopt review follow-ups

[bkaradzic-microsoft]

Summary

Follow-up for review comments on #1885, #1890, and #1897.

• Reject multisampled cube render targets in texture initialization and framebuffer creation instead of producing silently wrong pixels. Zero-fill every cube render-target face, matching WebGL texImage2D(..., null), and read all six faces before any clear.
• Replace the 240-callback validation cap with a 60-second elapsed deadline. Inspect every render pass the next frame can use, including outputRenderTarget.renderPassId, rig cameras, and scheduled custom targets. Enroll only utility layers that render automatically, and re-check shouldRender so enable/disable takes effect. Poll every associated scene before combining readiness.
• Describe _native.decodeMeshopt as a compatibility export. Pinned Babylon.js 9.21.2 and the current public MeshoptCompression implementation do not reference it.

Validation

node --check Apps/Playground/Scripts/validation_native.js, plus an in-process run of the scheduling self-check (output-target pass, rig cameras, dynamic utility-layer enable/disable, and non-short-circuiting scene polling).

The cube readback test is NativeEngineCubeRenderTargets.RejectsMultisamplingAndZeroFillsFacesBeforeClear. It was not executed in this environment.

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

Review effort: Lite
Findings: 2 · 4

Open (6)

• This loop has no overall timeout, so if the promise is never fulfilled (e.g., readback callback not… · New
• This multisampled-cube rejection only inspects colorTextures. If samples > 1 and… · New
• Using a plain object as a set (seen[id]) can collide with inherited keys (e.g., __proto__) and… · New
• The self-check runs unconditionally at script load and throws on failure, which can abort the… · New
• The self-check runs unconditionally at script load and throws on failure, which can abort the… · New
• The self-check runs unconditionally at script load and throws on failure, which can abort the… · New

What changed in this PR

Follow-up changes addressing prior review feedback around cube render targets, native validation readiness/convergence logic, and Meshopt export documentation.

Changes:

• Reject multisampled cube render targets and ensure cube render-target faces are zero-initialized.
• Make native validation convergence time-based (60s) and broaden readiness checks across render passes / cameras / utility layers.
• Clarify _native.decodeMeshopt as a compatibility export in code comments/docs.

File	Description
Plugins/​NativeMeshopt/​Source/​NativeMeshopt.cpp	Updates comments clarifying decodeMeshopt as a compatibility export.
Plugins/​NativeMeshopt/​README.md	Updates documentation of the compatibility entry point behavior/usage.
Plugins/​NativeMeshopt/​Include/​Babylon/​Plugins/​NativeMeshopt.h	Updates header comments to match current Babylon.js Meshopt behavior.
Plugins/​NativeEngine/​Source/​NativeEngine.cpp	Adds explicit rejection of multisampled cube render targets during init and framebuffer creation.
Documentation/​AddingNewValidationTests.md	Updates guidance to reflect new convergence and render-pass readiness behavior.
Core/​Graphics/​Source/​Texture.cpp	Zero-fills cube RT memory on creation to match WebGL semantics.
Apps/​UnitTests/​Source/​Tests.NativeEngine.CubeRenderTargets.cpp	Adds a unit test for multisample rejection + zero-fill behavior.
Apps/​Playground/​Scripts/​validation_native.js	Reworks convergence logic and render-pass enumeration; adds scheduling self-check.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

Review effort: Lite
Findings: 2 · 1

Open (3)

• hasPassId currently treats any non-null/undefined value (including NaN, objects, and strings)… · New
• The new multisample rejection for a cube depthStencilTexture attachment is not covered by the… · New
• The same error string is duplicated in multiple places in this file for the same condition. To… · New

Resolved since last review (6)

• This multisampled-cube rejection only inspects colorTextures. If samples > 1 and…
• This loop has no overall timeout, so if the promise is never fulfilled (e.g., readback callback not…
• The self-check runs unconditionally at script load and throws on failure, which can abort the…
• The self-check runs unconditionally at script load and throws on failure, which can abort the…
• The self-check runs unconditionally at script load and throws on failure, which can abort the…
• Using a plain object as a set (seen[id]) can collide with inherited keys (e.g., __proto__) and…