WebGL Render Warning if cacheAsBitmap is true

[jsfi]

If cacheAsBitmap is set to true on a container that is empty or has only invisible children, Chrome (v56) and Firefox (v49) log a WebGL warning.

Warning Chrome:

[.Offscreen-For-WebGL-0000005524CBB4F0] GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : glClear: framebuffer incomplete
[.Offscreen-For-WebGL-0000005524CBB4F0] RENDER WARNING: texture bound to texture unit 0 is not renderable. It maybe non-power-of-2 and have incompatible texture filtering.

Warning Firefox:

Error: WebGL: clear: Incomplete framebuffer: Status 0x8cd6: COLOR_ATTACHMENT0 has no width or height.
Error: WebGL: drawElements: Active texture 0 for target 0x0de1 is 'incomplete', and will be rendered as RGBA(0,0,0,1), as per the GLES 2.0.24 $3.8.2: The dimensions of `level_base` are not all positive.

Example: https://codepen.io/anon/pen/PWrXOB

Actual use-case: I have a container with n sprite elements. Depending on the state of the application I want to show 0 to n of the elements (by setting visible to true/false). After updating which sprites are visible, the container can be cached. If all elements are set to invisible the warnings appear.

My suggestion would be to extend the condition in
DisplayObject.prototype._renderCachedWebGL (and maybe in DisplayObject.prototype._renderCachedCanvas to be consistent) to include a check for width and height of the element.

if (!this.visible || this.worldAlpha <= 0 || !this.renderable)
{
  return;
}

to

if (!this.visible || this.worldAlpha <= 0 || !this.renderable || this.width === 0 || this.height === 0)
{
  return;
}

I can create a Pull request if you like the solution but I wanted to discuss it first as I am not sure if it's the only/best place for a fix.

[ivanpopelyshev]

better not use "width" and "height" , they are properties

[jsfi]

Is it better to use the bounds Object in DisplayObject.prototype._initCachedDisplayObject?

If the renderTexture is only created when bounds.width > 0 && bounds.height > 0, the cached sprite uses Texture.EMPTY as a fallback which doesn't trigger the warnings.

[ivanpopelyshev]

I'm not sure that bounds won't be changed later. Its better to add flag to cached object that renderTexture is empty, or may be reference Texture.EMPTY instead, and then check it in rendering method

[jsfi]

Maybe I understand something wrong, but aren't the bounds used to generate the texture?

https://github.com/pixijs/pixi.js/blob/dev/src/extras/cacheAsBitmap.js#L186

If either width or height are zero, the texture should be empty. So by checking width and height first, it is possible to step over lines 178 to 205 and create the sprite without texture (which will automatically use Texture.Empty https://github.com/pixijs/pixi.js/blob/dev/src/core/sprites/Sprite.js#L103)

[BossLevel]

Referencing my previously closed #3418 issue - it's just reared it's head again, and having read this, I can confirm that the relevant container has zero children at the time. I'd turned off cacheAsBitmap previously and this must have been what resulted in a resolution.

[neutrino84]

@ivanpopelyshev I was looking at the cacheAsBitmap implementation, and it seems bit heavy - would simply rendering to a texture do the trick? Cobbled something that sees to work, although, you need access the gl

[ivanpopelyshev]

@neutrino84 is does rendering into the texture and it works. what's the problem?

[jsfi]

Sorry for the spam, I didn't know all commits would be referenced.

Maybe some code makes it clearer what I think would be a good solution. Please ignore the first two commits.

[PixelsCommander]

Also see this problem on Intel HD graphics card. So it is hardware related, but let`s be fair, IntelHD is world most popular built-in chip.

[englercj]

Hey @jsfi, if you think you have a solution for this can you put in a PR that references this issue?

Thanks!

[ghost]

Whats the current update on this? I'm still getting this same bug, even on that codepen. I'm in a dilemma right now where I have to choose between setting cacheAsBitmap = false (with lower performance) or cacheAsBitmap = true (with these warnings)

[themoonrat]

Does anyone on this thread still get this error with the latest pixi? Thanks!

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.