Expose effective connection type to Cronet

net/nqe/effective_connection_type.h

Index: net/nqe/effective_connection_type.h
diff --git a/net/nqe/effective_connection_type.h b/net/nqe/effective_connection_type.h
index 6160a4d69bd6756a39c13070ec2ae181f8c33efe..4fa8c58769c424455a8c98c3f045f1b51afa165c 100644
--- a/net/nqe/effective_connection_type.h
+++ b/net/nqe/effective_connection_type.h
@@ -15,6 +15,12 @@ namespace net {
 // EffectiveConnectionType of a network is independent of if the current
 // connection is metered or not. For example, an unmetered slow connection may
 // have EFFECTIVE_CONNECTION_TYPE_SLOW_2G as its effective connection type.
+// An invalid Java prefix to strip is specified to prevent the Java class
+// generator from automatically stripping off the common prefix
+// ("EFFECTIVE_CONNECTION_TYPE_").
+// GENERATED_JAVA_ENUM_PACKAGE: org.chromium.net
+// GENERATED_JAVA_CLASS_NAME_OVERRIDE: NetworkEffectiveConnectionType

[xunjieli, 2016/08/01 21:25:09]

So this will generate a NetworkEffectiveConnectionType.java file which includes
the enums as static ints. 

I think it will be confusing to have both NetworkEffectiveConnectionType.java
and EffectiveConnectionType.java. Our consumers will see both.

[tbansal1, 2016/08/02 19:03:11]

This is now very similar to how C++ enum LoadState is converted to Java IntDef
Status.

An alternate approach is to expose these only for testing (Like I was doing in
the previous patch). Then, add a test which ensures that the C++ enum and the
Java IntDef remain synchronized. Then, the production code would not specify the

ChromiumEffectiveConnectionType.java as a dependency, and consumers would only
see one of the two classes.

+// GENERATED_JAVA_PREFIX_TO_STRIP: PREFIX_NOT_PRESENT_
 enum EffectiveConnectionType {
   // The connection types should be in increasing order of quality.
   EFFECTIVE_CONNECTION_TYPE_UNKNOWN = 0,