Oli Scherer

can you elaborate? Maybe post the error output. Is it spurious and only happens sometimes?

Oh. I think it's fine to simply replace the output with `Some("".to_string())` (so fixing the test, not the code)

Another use-case from irc by @alteous > I would like to obtain a maximum integer value from a tree of structures. The context is these values are offsets into arrays,...

Have you checked whether building with mir optimizations makes a difference? They run on generic code, instead of on the monomorphized versions

Can't we just mod the normal writer to always do that? It's not like one can write code accidentally that only works in the presence of paritial writes

Yea attributes are a weird thing. I'm not sure how to handle them. We could rename fields to something like `%foo`, where the `%` denotes that the field is meant...

> Could avoid most annotations by using something like attr_width vs sub_item by default, but maybe that will be even more verbose? You mean to require adding `%` for non-attributes?

hmm.. I don't like that, because it exposes xml stuff to users of the struct who don't even care about xml

> I think an annotation would be a good choice. Maybe #[xml(attribute)]. Serde has no support for such features. `serde-xml` is just a `serde` format, we can't create features that...

We'll also need to enforce that attribute-fields come before regular fields, because the serializer cannot look ahead. This is probably only doable with a runtime error though.