Jim Blandy

@brendan-duncan wrote: > For developers like us who don't write WGSL directly, marking WGSL blocks as "unsafe" or "uniform" isn't really an option. Like Kelsey, I'm surprised by this. Couldn't...

Rust uses `mod`, `pub`, and `use`, with `as` as a secondary keyword. ``` mod slurve { pub fn knuckleball() { ... } fn knurl() { ... } } use slurve::knuckleball...

Whoa, that is not what I expected. Even so - I would expect all WGSL compiler back ends to already have renaming systems, for the cases where WGSL permits identifiers...

(Feel free to remove the 'editoral' label - I did think it was non-controversial, but now I'm not so sure.)

I'm mildly disappointed, but let's make this issue about clarifying the spec one way or another. @dneto's suggestion seems fine.

Supporting such declarations would be a backwards-compatible change, so it's fine with me to put this off until after V1.

Here's how I was thinking it could work (I haven't actually tried this yet): 1) For the first pass, we parse mostly as we do now, except that identifiers in...

You should use `Arena`, `UniqueArena`, and `Handle` for the unresolved representation. That is one of the tricks Naga uses for performance: broadly speaking, caches and prefetchers are much happier with...

Using arenas instead of trees for expressions should reduce the stack depth, though, because you can just iterate over the arena directly. And with a little care we could avoid...

`stacker` sounds 1) amazing and 2) wildly unsafe and definitely not something we want to ship in a browser. So that's one easy solution out the window. I think stack...