[tlively]

Update parsing of binary type sections to use TypeBuilder to support uses before
definitions. Now that both the binary and text parsers support out-of-order type
uses, this PR also relaxes the logic for emitting types to allow uses to be
emitted before definitions.

[kripken]

Nice!

lgtm aside from the one question.

[kripken]

why not 0? (is -1 guaranteed to segfault on use like 0 is?)

[kripken]

Actually, is this needed? It would be better to avoid allowing an invalid HeapType at all. It seems like getBasicHeapType returns a boolean already, so we know if a value was found or not. Given that, the HeapType could be initialized to a valid value (say, anyref, doesn't matter) before the call? (There is a risk to doing that as well, but with a good comment there I think it's less risky than a general change to allow invalid HeapTypes to be declared as easily as HeapType bad;).

[tlively]

Sure, I can have the default constructor just choose an arbitrary BasicHeapType. FWIW, address -1 should segfault on all reasonable platforms (as long as you're not in kernel mode or in WebAssembly or something 😉).

[kripken]

Side note, @jakobkummerow and I happened to look at a wasm file with recursive types now (testcase for the names section), and I noticed that it crashes with this PR,

==580908== Stack overflow in thread #1: can't grow stack to 0x1ffe801000
==580908== 
==580908== Process terminating with default action of signal 11 (SIGSEGV)
==580908==  Access not within mapped region at address 0x1FFE801FF0
==580908== Stack overflow in thread #1: can't grow stack to 0x1ffe801000
==580908==    at 0x507BBEC: std::_Function_handler<void (unsigned long), wasm::(anonymous namespace)::Canonicalizer::getOrderedItems()::{lambda(unsigned long)#1}>::_M_invoke(std::_Any_data const&, unsigned long&&) (in /home/azakai/Dev/2-binaryen/lib/libbinaryen.so)

(Infinite recursion?)

That wasm has something like this:

// type 0: struct $StrA (field ($byte i8) ($word i16) ($pointer (ref $StrB)))
// type 1: struct $StrB (field ($next (ref null $StrA)))

name-section.wasm.gz

[tlively]

Yeah there are going to be lots of infinite recursions for recursive types right now, but I'm surprised that this didn't trigger an assertion in the TypeBuilder first.