Fix a panic in table-ops translation

[alexcrichton]

This fixes an issue where ensure_inserted_block() wasn't called before
we do some block manipulation in the Wasmtime translation of some
table-related instructions. It looks like ensure_inserted_block() is
otherwise called on most instructions being added, so we just need to
call it explicitly it seems here.

Closes #2347

[alexcrichton]

I'm a bit curious why the fuzzers haven't found this one yet, @fitzgen do you know if wasm-smith will create an if else end empty block like that?

[fitzgen]

wasm-smith definitely should be able to generate if; else; end sequences. It might be somewhat low probability, since all of the following need to happen:

• We need to choose to generate a stack neutral if control frame (1 out of 400 chance to generate a stack neutral type; not sure on probability to generate any if in the program at all)
• We need to immediately choose an else instruction (1 out of ~250 or so; not sure how many instructions we support these days)
• We need to then immediately choose and end instruction (1 out of ~250 again)

so that's a 1 out of ~25,000,000 chance, ignoring running up against the entropy limit. Which seems like reasonable odds to hit fairly quickly, given that our fuzzers are running continuously...

So in conclusion: I'm not sure what's up, and I would have expected this to have been triggered earlier.

[fitzgen]

(Updated comment due to accidental early submit)

[fitzgen]

I think you need to split the new test file into one file for each module, which is why the tests are failing I think.

[alexcrichton]

Bah I needed to touch build.rs to actually run the tests locally, I got the order of operands to table.set wrong by accident.