Relax render pass color_attachments validation#2778
Conversation
3233906 to
7aebf7d
Compare
jimblandy
left a comment
There was a problem hiding this comment.
The rename looks good - could we split that out into its own PR?
However, the WebIDL and the spec language make it clear that colorAttachments can be an array with holes, like [x, null, y], so it really does need to be an Cow<'a, [Option<RenderPassColorAttachment>]> to match the spec.
1f2d807 to
823bd0a
Compare
|
Tested color_attachments: Since a new version of the aligned WebGPU spec TypeError: GPUDevice.createRenderPipeline: Missing required 'format' member of GPUColorTargetState.
__wbg_createRenderPipeline_9112d655cc89bd45 http://localhost:8000/skybox2.js:599 |
jimblandy
left a comment
There was a problem hiding this comment.
Thank you so much for undertaking this! That turned out to be a massive change.
…sses, and render bundles
823bd0a to
74ca2a3
Compare
jimblandy
left a comment
There was a problem hiding this comment.
This looks great - thank you!
The only thing that still concerns me is the handling of rtv_pool. Based on discussion in Matrix, I think we can avoid having to share rtv_pool altogether, by allocating the null handle we need earlier, and just passing it in to command buffer creation. See this comment.
74ca2a3 to
c49c13e
Compare
c49c13e to
02da0b0
Compare
* upstream GPUAutoLayoutMode * clean up symbols and fix miscalled op * fix #2778
Connections
#2469
Description
I tried
Option-al all of these, but that makes the code changes look not so good.Since
&[]can expressNone, is it still necessary to useOption?Testing
Tested locally