Proposal: enums in INIT macros always use Undefined

[kainino0x]

The actual default may be dependent on where the struct is used, or could be changed from a trivial default to a non-trivial default in the future.

The Undefined enum values (or WGPUCompositeAlphaMode_Auto) are defined to always do the defaulting inside the API regardless of whether it's a trivial default or not.

All enum fields (searched for type: enum\..*\n\s*default: ) now use Undefined(/Auto) or *NotUsed. (except for #445)

I also added documentation to all the ones I changed, explaining the default. All of the others are either optional or required in the JS spec, so should be fairly clear.

I might turn the defaulting/optional/required docs into something more generated later, but for now this does what we need.

[eliemichel]

If I understand correctly, setting e.g. WGPUBlendComponent to Undefined will have the same effect as setting it to Add (in the current stats of the API), so it is preferred to init it to Undefined.

I see the point of automatically handling future updates of the API, but on the other hand it makes the header much less self-contained (and knowing the default values is one of the main reasons why I always have to look up the docs).

I'm tempted to suggest we adopt a soft trade-off where the default value is in effect Undefined but the docstring explicitly tells what non-undefined value this corresponds to. We're back with the risk of doc being out of sync with the actual behavior but at least it solves the issue of expressing non-trivial defaulting scheme.

We could phrase the doc such that we invite one to check out a reliable source of info about where to find up-to-date information about the defaulting mechanism.

[kainino0x]

@eliemichel Sorry - I completely forgot to push my second commit before opening this PR (which I was referring to in my PR summary). It adds documentation for all of the defaulting. WDYT?

[eliemichel]

Solves my concern indeed!

[kainino0x]

Landing tentatively - pretty sure this is what we'll settle on, but we'll !discuss to verify.

[kainino0x]

Dec 9 meeting:

• KN: (summary)
• CF: Don't love it but seems fine. Prefer not to lose documentation
• KN: Agree, I added documentation for all of these in the PR.