Envoy test passed: envoyproxy/envoy#17547

What's the reasoning behind this change? Do you need it for something or is it a generic cleanup?

It's fine in principle, but I'm asking because the next iteration of the ABI is most likely going to explicitly include handles to host resources (i.e. context) in parameters, at which point we might need to revert this change (I'm not 100% sure yet, since it's not fully fleshed out).

This is the general cleanup.

What do you mean by explicitly include handles to host resources (i.e. context) in parameters? The host functions in ABI, for example proxy_get_header_map_pairs will have the signature like (*void, map_type, return_ptr, return_size)? I think this change is about the interface on the host defined functions in C++ and the host defined functions in the ABI, and I don't really understand why do we need to revert if the ABI changes. Sorry if I'm missing something.

If this is about context_id as I see in https://github.com/WebAssembly/WASI/blob/b35c78afd6224e5d27484c7613ca0fa46d9ac4c1/meetings/2021/presentations/2021-04-22-PiotrSikora-Proxy-Wasm.pdf, then the revert won't be needed because we can just treat it as Word like we do for other args now.

If this is about context_id as I see in https://github.com/WebAssembly/WASI/blob/b35c78afd6224e5d27484c7613ca0fa46d9ac4c1/meetings/2021/presentations/2021-04-22-PiotrSikora-Proxy-Wasm.pdf, then the revert won't be needed because we can just treat it as Word like we do for other args now.

It is. My point was that with that change we could simply replace void *raw_context with Word context_id, whereas if we remove it in this PR, then we'd need to re-add in the future, which seems a bit inconvinient... but not too much of an issue, so let's ignore it after all. Sorry for the noise!