Jonathan Wight
Jonathan Wight
Hi Adar, that's really interesting. I've been seeing this issue myself a little when switching targets in a project. Next time I encounter it I'll see if this fixes it...
I have a GitHub action failing with this issue and setting ENABLE_USER_SCRIPT_SANDBOXING to either NO or YES seems to make no difference.
Looking at the errors generated above (and in my failing GitHub action) and the output of `xcrun metal` I wonder if we can use ` -fmodules-cache-path` to set the cache...
I merged this PR in today. It's been working great for me and has solved the issue for me at least.
It's unlikely I'm going to be able to address this any time soon. I am happy to take pull requests though… Might be an easy fix?
Is there a crash? Or does it just render wrong? Other models have non-nil byte strides… I'll try and render your cube in Blender see what it's supposed to look...
I've pushed a change to main that I think might fix this. Assuming you were using the RenderKit code to convert GLTF to RenderKit models there was definitely an issue...
Yup. It's not really usable with current output. Small changes add a lot of extra lines to output.
Yup. Same - tracking issue.
Yeah latest version here. Would extra logging help at least provide any insight into what's causing it? Are you just displaying the error string some HTTP service gives you are...