Typed continuations: nocont and cont basic heap types

[frank-emrich]

This PR is part of a series that adds basic support for the typed continuations/wasmfx proposal.

This particular types adds cont and nocont as top and bottom types for continuation types, completely analogous to func and nofunc for function types (also: exn and noexn).

[tlively]

The isFunction / isSignature distinction is definitely something I would like to make better, perhaps by removing isFunction, but I haven't looked into it in detail. We also don't have different versions of isStruct or isArray that include the struct and array heap types, so we're already inconsistent about the API structure. All this is to say I think it would be fine (and even preferable) to undo the renaming of isContinuation.

[frank-emrich]

Fair enough, I'm happy to keep the old name. I've undone the renaming and updated the PR description.

[tlively]

LGTM, but for more test coverage it might be a good idea to add continuation types to TestHeapTypeRelations in test/gtest/type-builder.cpp as well.

[frank-emrich]

Good call, I actually caught a place I had forgotten to update this way!

[tlively]

Excellent!

[frank-emrich]

Just FYI, this was the last missing piece to add basic support for the core of the proposal to binaryen. Only resume.throw and barrier instructions are missing, which aren't a high priority for us at this point.

I'd therefore prioritize implementing validation for the already added instructions first. Can the new wat/wast parser now be used for spec tests, or is that going to be the case soon-ish? I'm asking because I'm hoping to re-use our existing spec tests when adding validation, rather than having to transform them into binaryen's text format.

[tlively]

No support for running spec tests with the new parser yet, but I hope to be done with that within a couple weeks.

[tlively]

@frank-emrich, the new parser is now used for top-level modules in wasm-shell, so you should try running your spec tests now. It's possible you'll run into some other limitations, but at least there should be fewer now. I plan to keep working toward being able to run the full upstream test suite.

[frank-emrich]

That's great, thanks for letting me know!