Update the memory abbreviation in the text format

[BilelGho]

The current proposal for the extension of the memory abbreviation does not take into consideration the memory64 feature and the existence of an addrtype:

The memory abbreviation is extended to allow an optional page size as well:

'(' 'memory' id? mempagesize? '(' 'data' b_n:datastring ')' ')' === ...

Is mempagesize? supposed to be before or after addrtype? ?

[tommie]

https://webassembly.github.io/memory64/core/text/modules.html#text-mem-abbrev

The address type is i32 or i64 so the ambiguity is cosmetic in WAT. However, perhaps it would make sense to wrap the mempagesize in a tagging tuple, just like data.

(The binary format docs place addrtype first, so I'd propose consistency with that.)

https://github.com/WebAssembly/custom-page-sizes/blob/main/proposals/custom-page-sizes/Overview.md#binary-encoding

[fitzgen]

FWIW, our implementation in wasm-tools doesn't actually implement that abbreviation as an oversight. I guess we need a test that uses this syntax.

(The binary format docs place addrtype first, so I'd propose consistency with that.)

This makes sense to me.

[fitzgen]

FWIW, our implementation in wasm-tools doesn't actually implement that abbreviation as an oversight. I guess we need a test that uses this syntax.

Nevermind, I spoke too soon, we do implement it and we do in fact parse it after the index type:

https://github.com/bytecodealliance/wasm-tools/blob/bc0a78578a764b3f43489e210066bc7ea2449ac5/crates/wast/src/core/memory.rs#L68-L73

[conrad-watt]

quick additional check - where does the shared annotation from threads go in relation to this?

EDIT: from WebAssembly/threads#226, did we just never define this?

[fitzgen]

EDIT: from WebAssembly/threads#226, did we just never define this?

Yeah, it seems like that was never defined. wasm-tools at least doesn't support shared in that abbreviation.