Regression in r127dev MultiRenderTarget GL_INVALID_OPERATION: Blit feedback loop: the read and draw framebuffers are the same.

[arpu]

r127dev
https://raw.githack.com/mrdoob/three.js/dev/examples/webgl2_multisampled_renderbuffers.html

works on stable r126
https://threejs.org/examples/?q=webgl2_multisampled_renderbuffers.#webgl2_multisampled_renderbuffers
To Reproduce

Live example

Expected behavior

should work on r127dev

Screenshots

Platform:

• Device: [Desktop]
• OS: [ Linux]
• Browser: [Chrome 89]
• Three.js version: [dev]

[Mugen87]

Good catch! It seems the E2E tests are not yet able to detect this issue.

[arpu]

i see this problem on testing MultiRenderTarget on webgl2 for MSSA >4 would be nice to add this to the example? with gl.MAX_SAMPLES

[mrdoob]

i see this problem on testing MultiRenderTarget on webgl2 for MSSA >4 would be nice to add this to the example? with gl.MAX_SAMPLES

Would you like to do a PR for it?

[arpu]

i have some problems on this, i always get 0 Samples used in Spector.js

[arpu]

would be nice to get some hints how i can use this
testing on https://jsfiddle.net/xgw4705v/13/

[arpu]

looks like i need this #16895 to get this done

[mrdoob]

Feel free to do a new PR to replace #16895 👍