Naming of segments vs the table they belong to

[sbc100]

Perhaps I'm missing some context but its seems like previouly valid wat files are no longer valid under this proposal because the meaning on the $name in the top level elem segments changed:

(elem $foo (i32.const 0))

Currently $foo names a table, but with this proposal $foo now names the element segment itself.

This change be seen in the changes to test/core/elem.wast compared to upstream.

This (AFAICT) s causing some problems in in wabt today: WebAssembly/wabt#1193.

[rossberg]

Yes, I think you're right. Technically, this is indeed a breaking change to the text format (the same applies to data segments, FWIW). However, it seems fairly minor in practice, since the name was pretty much useless before, so was hardly used, and moreover, the only valid modules that actually break are ones that have multiple element segments with explicit named table references.

Is it worth backporting this change to the MVP to avoid the wabt hack?

[sbc100]

I ended up ignoring elem names unless bulk memory is enabled: WebAssembly/wabt#1193.

It seems to work OK for now. If this seems like a reasonable approach we can close this issue.