Thomas Lively

https://github.com/rust-lang/rust/pull/73526#issuecomment-675795604 is a timely example of how this would be beneficial to Rust.

The first version after the branch was 1.39.20. I had to figure that out to make the Rust CI use that version after they switched to using LLVM 11.

This logic is responsible for skipping the reuse of an existing table and will cause a new table to be created on demand: https://github.com/WebAssembly/binaryen/blob/main/src/ir/module-splitting.cpp#L175-L178 The fuzzer should make sure that...

@dschuff added the special casing of emscripten tables in https://github.com/WebAssembly/binaryen/pull/7050. It sounds like Emscripten's loading code couldn't handle the second table at the time. @aheejin, is this something we could...

Yes, it is production-ready. There unfortunately are not great docs for it, but you can see how to create and mount the backend in the [tests](https://github.com/emscripten-core/emscripten/blob/bd4676aa00785a6a1fad4a54a168a7b594540d10/test/wasmfs/wasmfs_opfs.c#L28-L33).

I'm not sure what promises you're referring to. Are you using `-sASYNCIFY=1` in your normal builds? IIRC, WasmFS doesn't support ASYNCIFY=1. @brendandahl, can you speak to how experimental the JSPI...

Are you able to reproduce this by calling wasm-opt directly? If so, would you be able to profile the wasm-opt run on Windows and share the results?

That could make sense. I'm not seeing very many occurrences of branch hint bugs, so keeping it enabled doesn't seem too disruptive, either.

Also pages 293-304

Adding new embedding functions to fix this violation of the embedder API abstraction sounds good to me.