Prevent errors when communicating with iOS webview

[domchristie]

To successfully communicate with the iOS view, messages need to be of a type supported by the structuredClone algorithm. The new TransferableVisitOptions type ensures that VisitOptions passed to the adapter conform to this type. The StructuredCloneValue type has been created based on https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm#supported_types

This should hopefully prevent future errors such as #720

---

An alternative to this might be that we become stricter with the VisitOptions we pass around, and instead of using Partial use Pick and Omit?

/cc @kevinmcconnell

[kevinmcconnell]

This is a great idea! 👏 I really like the idea of using the compiler to enforce the safety of the options fields, rather than rely on a comment (or our memory).

My only thought (as mentioned below) is that I'd be tempted to tweak it so that adding a non-transferable option would cause a compiler error, rather than it being automatically sanitised. It just makes it more obvious to the person adding that field that that's happening.

[domchristie]

Just to add: this doesn't change the existing behaviour; it just adds type checking, and removes any properties that would cause a DataCloneError

[domchristie]

(And I'm not sure why the tests are failing 🤷‍♂️ They also fail for me locally on the main branch`)

[domchristie]

Just to add: this doesn't change the existing behaviour; it just adds type checking, and removes any properties that would cause a DataCloneError
(And I'm not sure why the tests are failing 🤷‍♂️ They also fail for me locally on the main branch`)

On second thoughts, this does alter the behaviour, because the options passed to Navigator#startVisit will be without referrer and visitCachedSnapshot. This is amended in 122fd26

[domchristie]

Superseded by #750