`IsComplete` may check `requires`, making checking before writing to a 0-initialized buffer difficult.

[jasongraffius]

In the C++ runtime if a type has a requires clause and that field is used in determining the size of the struct, IsComplete appears to check that requires clause. This means a generic approach of 0-initializing a buffer then checking IsComplete breaks for structs that have fields that exclude 0, for instance.

Workarounds involve either leaving it unchecked and manually confirming (defeating the purpose of checks like IsComplete) or splitting definitions into fixed-size portions and variable-sized portions.

We should evaluate whether IsComplete should ignore the requires for these fields, or if there's a good reason we need to ensure the consistency of these fields, potentially provide a verifier that ensures that all fixed-size, fixed-location portions of the struct are safe to write to, allowing the client code to call IsComplete once they've initialized those to ensure the entire struct is safe to write to.

[reventlov]

For some background, IsComplete() was intended to be used (and is used) to determine whether adding more bytes to the buffer could change the interpretation of the buffer -- literally, "does this sequence of bytes contain a complete (but possibly invalid) message, or are there more bytes remaining?" It seems like returning false for an unmet [requires] on a sizing field would break that.

I suppose it also has the overloaded meaning of "can you read the intrinsic size?" -- although IntrinsicSizeInBytes().Ok() (IIRC) should tell you that.