Tests fail with --no-default-features on published "serde_bytes" 0.11.3 crate about bytes

[dtolnay]

default-features = false is supported.

[dtolnay]

I would prefer to leave it. All the cfg features we use are additive, so if the crate compiles successfully with various feature combinations and passes tests with all features enabled then it's fine. Introducing cfg gunk into the test suite would make it less likely to cover what we want.

[kentfredric]

That sounds reasonable. Just a follow up though, is it fair to say that this means "default-features = false" is inherently unsupported? ( Or perhaps, unsupported unless your crate intends to also demand futures specifically selected by their consumers ).

If so, it may pay to put a cfg! bomb in there so it won't build without one of a minimal set.

( Otherwise, the existence of this bug will serve its own purpose if anyone else goes looking )