Thomas Lively

> If there is a switch handler, set the cache before resuming the continuation and clear the cache if the continuation returns. This would only be correct if the resumed...

FWIW, I don't think we expected the call_indirect issue before we made the change, but I don't think we would have done much differently if we had anticipated it (except...

In general, it would also be a good idea for engines to support encoding changes introduced by new proposals, even if they are configured not to support the new features...

To be clear, I'm only talking about proposals that have become part of the standard. The core spec tests need to be updated for such proposals anyway. Accepting new encodings...

I'm not suggesting that feature flags be removed, but only that the decoder (and no other part of the codebase) behave as though standardized features are always enabled. It's a...

FWIW, we're about to start importing the upstream spec tests into Binaryen as a submodule, so we'll start running into the same issues. Another way you can minimize pain here...

It could in principle, but that would be an expansion of its responsibilities and would make the abstraction fuzzier. I think this separation of concerns is nicer, but I can...

Ok, I looked into the performance overhead of this change and I think it is acceptable. In the very worst case where `--debuginfo` is used, but there are not actually...

Right, we currently do the walk whenever --debuginfo is used, and if there are no names to apply, it simply won't make any changes. I had been thinking that the...

I guess 6% is the relevant worst-case overhead we've seen, but I think it's also notable that on real modules we've also seen 0.2%. This shouldn't be any different in...