Vecvec

This fixes an annoying access violation termination (that I thought was a driver bug) on program exit after a present.

Sorry for a slow response, in reply to your first comment, I think you raise some good points, and I agree that an API re-design should not be attempted in...

@JMS55, @daniel-keitel While working on a new example, I ran into a issue with building that causes a vulkan device lost in a later call (but only if the struct...

I'm not sure keeping the tlas alive is necessary for this issue, I think that line (and the corresponding one in render.rs at line 1516) can be swapped out to...

looking into the Vulkan issue, Nsight runs it unless you disable Device adress C++ support, and if disabled it will warn > ID,,Origin,Source,Message 25,,Target,NVIDIA Nsight Graphics,"Buffer device address was obtained...

@JMS55 Yes, though that might need some extra work, I will design an api suggestion and post it on the ray-tracing issue (as it may need a follow up PR).

>Tests fail now though, something about resource tracking stack overruns I've been working on fixing use after free crashes (and merged trunk while doing so), and I've encountered a similar...

That's great! What did you change? Also I think that that validation error is similar to #5373, same submitted usage though different prior usage. Edit: actually it has different write...

I've found the issue, turns out we need to call transition buffers on the tlas instance too (seems like a preexisting issue in ray-tracing).

I've opened atlv24/wgpu#1 for the write after write fix.