The failure dependency doesn't do much here, and is a bit heavy because of the transitive dependency to syn. Removing it reduces the release build time from 13 to 4 seconds.
failure also brings some interoperability problems with crates that don't use it.
I can submit a PR to replace it with a hand-rolled implementation, but it's technically a breaking change. Do you have anything against it? And if we introduce a breaking change now, can we also update to the 2018 edition and start using dyn Trait to get rid of some compiler warnings?
The
failure
dependency doesn't do much here, and is a bit heavy because of the transitive dependency tosyn
. Removing it reduces the release build time from 13 to 4 seconds.failure
also brings some interoperability problems with crates that don't use it.I can submit a PR to replace it with a hand-rolled implementation, but it's technically a breaking change. Do you have anything against it? And if we introduce a breaking change now, can we also update to the 2018 edition and start using
dyn Trait
to get rid of some compiler warnings?