Fix/webgpu crash pixel density

[saurabh24thakur]

Fixes #8456

Description

Fixes a crash when using the WebGPU renderer where calling pixelDensity() immediately after setAttributes() (or any other function that triggers an asynchronous context reset) would result in TypeError: Cannot read properties of undefined (reading 'createTexture').

The Issue

In p5.js 2.0, setAttributes() triggers an asynchronous reset of the rendering context via _resetContext(). Because core functions like pixelDensity() are synchronous, they can trigger a resize() (and subsequently _updateSize()) on a new renderer instance before its WebGPU device has been successfully initialized.

Changes

1. RendererWebGPU: Added guard clauses to _updateSize(), clear(), and clearDepth() to return early if this.device is not yet defined. Since _initContext() calls these methods upon successful initialization, the state is eventually synchronized correctly once the device is ready.
2. Renderer3D:
   • Updated _resetContext() to await the new renderer's contextReady promise before replacing the existing renderer on the p5 instance. This ensures that the sketch's _renderer property always points to a valid, initialized renderer.
   • Ensured the new renderer inherits the pixelDensity of the renderer it is replacing to avoid size-mismatch warnings in the console (e.g., depth-stencil attachment size mismatches).
3. Tests: Added a new unit test in test/unit/webgpu/issue_repro.js that specifically reproduces this race condition.

Checklist

• I have read the CONTRIBUTING document.
• My code follows the style guidelines of this project.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• My changes generate no new warnings.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.

[davepagurek]

Timeouts in tests are fairly brittle, we can use expect(() => { ... }).not.toThrow() or something like that as an alternative.

[davepagurek]

Can we use p.WEBGPU rather than the value of the constant?

[davepagurek]

I'm not 100% sure that we'll do something with this, can we try changing antialias as a more normal example?

[davepagurek]

A new file called issue_repro.js is not really readable for future contributors trying to understand the structure of our code and tests. Could this just be a test in the existing WebGPU renderer tests?

[davepagurek]

To clarify, is this added just to methods that get called automatically when you reinitialize the context?

[davepagurek]

Thanks for working on this! Left a few comments, let me know what you think!

[saurabh24thakur]

Thanks for the review, @davepagurek Those points make total sense. I'll move the test into the main WebGPU suite and swap out the setTimeout for a cleaner expect().not.toThrow() assertion. I’ll also update the test to use p.WEBGPU and antialias as you suggested. I'll double-check the guard clause scoping and push the updates shortly!