api: remove Future.Err()

The method causes an improper error handling because it returns an error
only from a client side. Therefore, it is not enough to simply check
the error to find out that the request was not completed.

A user should check an error from `Future.Get()` or
`Future.GetTyped()`. In addition, the user can find out whether there
was an error without decoding the response body with
`Future.GetResponse()`, see `ExampleErrorNo`.

Author: oleg-jukovec

CHANGELOG.md

@@ -127,6 +127,7 @@ Versioning](http://semver.org/spec/v2.0.0.html) except to the first release.
 - `Schema` field from the `Connection` struct (#7)
 - `OkCode` and `PushCode` constants (#237)
 - SSL support (#301)
+- `Future.Err()` method (#382)
 
 ### Fixed
 

README.md

@@ -348,6 +348,9 @@ for an `ops` field. `*Operations` needs to be used instead.
 * `Future` constructors now accept `Request` as their argument.
 * Methods `AppendPush` and `SetResponse` accepts response `Header` and data
   as their arguments.
+* Method `Err` removed because it provokes an invalid error checking. You need
+  to check an error from `Get`, `GetTyped` or `GetResponse` with an addition
+  check a value of `Response.Header().Error`.
 
 #### Connector changes
 

future.go

@@ -266,11 +266,3 @@ func (fut *Future) WaitChan() <-chan struct{} {
 	}
 	return fut.done
 }
-
-// Err returns error set on Future.
-// It waits for future to be set.
-// Note: it doesn't decode body, therefore decoding error are not set here.
-func (fut *Future) Err() error {
-	fut.wait()
-	return fut.err
-}