Conrad Watt

IIUC, the complication here in comparison to `memory.fill` would be the potential for a racing thread to observe the unsanitised surrogate before it's overwritten. We could write the specification for...

Currently the specification is just unordered byte writes (it would be equivalent to have one big tearing unordered write). This is a deliberate choice which means that any racing read...

> So the intended implementer guidance is "reason at the level of the ISA"? Yes, for maximum performance. `Relaxed_Memmove` would also be a conformant but potentially less ambitious way to...

Is the public "review" mentioned above the same thing as the "exclusion opportunity" triggered by new CRs? Interpreting the [above FAQ](https://sideshowbarker.github.io/w3c-faq/#how-can-i-make-a-living-standard-from-my-spec), new pushes to W3C TR don't trigger new exclusion...

I should say explicitly that thinking through the implications of a separate continuation closure type, things seem to get quite messy (e.g. with separate resume variants for each type), which...

Ah interesting, I think this answers my question about how the unboxed representation would work (assuming the stack resource is still a separate allocation). I didn't realise `cont.bind` would invalidate...

A deterministic replay mode for arbitrary Wasm programs with shared-memory concurrency would take a lot of engineering to support in a production engine, so I don't think we can mandate...

> Although it is necessary for the target continuation's return type to match the return type expected at the handler and communicated via the tag, it is not necessary for...

Also noting that @rossberg's original comment (https://github.com/WebAssembly/stack-switching/issues/61#issuecomment-2222629632) gets this right, and I've been overzealous in adding subtyping :) I think the subtyping on the "switched-to" continuation is ok, but the...

Actually, even this might not be enough, because currently the typing rules for `switch` handlers that I wrote down allow the recorded return type to be a subtype of the...