Open reuvenpo opened 6 years ago
It's documented, but admittedly not in a place where you'd look. https://github.com/Arnavion/derive-error-chain/blame/741135fc0ef00b204e6661eb283ead6e28fc7fb3/CHANGELOG.md#L60
Making error-chain a dependency of this crate will not work. It needs to be a dependency of the crate where the macro is invoked.
I see... I think it's a good idea to put it somewhere near the top of the documentation. It's a pretty important, non-trivial detail (and i spent an hour before i figured it out...)
Is it possible/desirable to lose that dependency? or would that both break compatibility with error_chain
and be too much of a hassle?
(of course, the saner alternative is to just document the dependency :P)
Using error-chain's traits is a deliberate decision. https://github.com/Arnavion/derive-error-chain/issues/5
That makes sense.
(Thanks for the cool package, btw :) )
It is undocumented that using a given version of
derive_error_chain
requires the use of a specific version oferror_chain
.For example, to use
derive-error-chain = "=0.11.2"
, one also has to specifyerror-chain = "=0.11.0"
inCargo.toml
. otherwise, one gets unexpected errors from inside the procedural macro (Exact errors depend on specific version combination).This should either be documented, or otherwise include
error-chain
as an explicit dependency (since one doesn't even need toextern crate error_chain
in one's crate root)