Rerun bootstrapping, and use a consistent encoding method in proto-le…

…ns-protoc. (google#290)

Previously we were always using reflected encodings in proto-lens-protoc.
After recent changes, the generated encodings are good enough that we can
use them to decode DescriptorProto.  So, this change makes us use the generated
encodings (for better testing) when the flag is set.

app/protoc-gen-haskell.hs

@@ -14,13 +14,11 @@ import Data.Monoid ((<>))
 import qualified Data.Set as Set
 import qualified Data.Text as T
 import Data.Text (Text, pack)
-import Data.ProtoLens (defMessage)
+import Data.ProtoLens (defMessage, decodeMessage, encodeMessage)
 -- Force the use of the Reflected API when decoding DescriptorProto
 -- so that we can run the test suite against the Generated API.
 -- TODO: switch back to Data.ProtoLens.Encoding once the Generated encoding is
 -- good enough.
-import Data.ProtoLens.Encoding.Bytes (runParser, runBuilder)
-import Data.ProtoLens.Encoding.Reflected (parseMessage, buildMessage)
 import Lens.Family2
 import Proto.Google.Protobuf.Compiler.Plugin
     ( CodeGeneratorRequest
@@ -51,9 +49,9 @@ main :: IO ()
 main = do
     contents <- B.getContents
     progName <- getProgName
-    case runParser parseMessage contents of
+    case decodeMessage contents of
         Left e -> IO.hPutStrLn stderr e >> exitWith (ExitFailure 1)
-        Right x -> B.putStr $ runBuilder $ buildMessage $ makeResponse progName x
+        Right x -> B.putStr $ encodeMessage $ makeResponse progName x
 
 makeResponse :: String -> CodeGeneratorRequest -> CodeGeneratorResponse
 makeResponse prog request = let