Suggest #[serde(bound="")] when encountering failure to infer bounds

[graydon]

I filed a compiler bug on this too (rust-lang/rust#73082) but it would have been helpful to see suggested in the serde docs that when encountering a type annotations needed / cannot infer type for type parameter error when attempting to derive(Deserialize), the solution isn't always to add an annotation or to specify a Deserialize bound: in the case I ran into today (where the type parameter it was struggling with already implements DeserializeOwned), the solution is to annotate with #[serde(bound="")], disabling serde's attempt at inferring a bound entirely.

I can't tell whether the right place to discuss this would be in https://serde.rs/container-attrs.html or https://serde.rs/attr-bound.html or some other place, the nearest hint I found on the matter was on stackoverflow (which I've added an extended answer for, in case anyone stumbles on it again: https://stackoverflow.com/a/62239787/1669902)

[swfsql]

Related issue: #1296

[dtolnay]

I think it's unlikely that adding information about this to the documentation would be a net benefit. In general, with more documentation of edge-cases, people read even less of the higher-value documentation, because the overall relevance is diluted. I think this specific issue comes up seldom enough that Stack Overflow can be sufficient coverage.

In your case, going by the rust-lang/rust issue, the problem was a consequence of rust-lang/rust#41617, which I am still hopeful will get fixed in rustc eventually.