Results 545 comments of David Neto

The rule that vertex shader must emit `position` is in the function Restrictions section. We should make that consistent with the compute shader restriction.

I'm not 100% clear if this is an array named "vector" or an access on a vector value. Let's make this concrete: ``` fn foo() { var A: array; var...

> So the details need to be refreshed with this in mind. I think indexing with const expr outside of bounds should be allowed to be erroring at compile time....

Sorry, I should have expressed my preference: vector indexed via const_expr out of bounds --> shader creation error array indexed via const_expr out out of bounds -> dynamic error (keeps...

Consensus from 2021-06-08 is: * treat it as a dynamic issue rather than static. * still can use a clarification in the spec that even though all the index and...

> This means, any access to an array is a dynamic access, if we go this route. In turn, that means by-value arrays can't be indexed in WGSL Not quite....

> At some point, didn't we have a proposal for mat.x? I thought @kvark had an open issue for it, but I can't find it.

After further discussion, we don't have consensus inside Google. (Sorry) * @alan-baker still thinks its better for the spec to have a consistent rule for invalid array accesses. * Will...

#3400 landed the change to make this case an error. Closing.