refactoring Android Webview JNI_OnLoad interface

base/android/library_loader/library_loader_hooks.cc

Index: base/android/library_loader/library_loader_hooks.cc
diff --git a/base/android/library_loader/library_loader_hooks.cc b/base/android/library_loader/library_loader_hooks.cc
index 97aec8c56ad5c9c9cc758fc74f7ab70c6473580c..e4451156a19de2fb09a41c91de1d6745bec34c71 100644
--- a/base/android/library_loader/library_loader_hooks.cc
+++ b/base/android/library_loader/library_loader_hooks.cc
@@ -123,9 +123,10 @@ void LibraryLoaderExitHook() {
 }
 
 bool RegisterLibraryLoaderEntryHook(JNIEnv* env) {
+  // TODO(michaelbai): Remove this after refactoring all dependents.
+  InitAtExitManager();
   // We need the AtExitManager to be created at the very beginning.
-  g_at_exit_manager = new base::AtExitManager();
-
+  DCHECK(g_at_exit_manager);
   return RegisterNativesImpl(env);
 }
 
@@ -142,5 +143,9 @@ LibraryProcessType GetLibraryProcessType(JNIEnv* env) {
       Java_LibraryLoader_getLibraryProcessType(env));
 }
 
+void InitAtExitManager() {

[Torne, 2015/03/11 18:08:29]

Why do we need to InitAtExitManager in this way at all? Can't it be done during
the regular init for base? I assume something that's happening during
registerjni requires it; what is that? SHould it be there?

[michaelbai, 2015/03/11 18:29:10]

As I said in my previous comment, I am refactoring other components which
initialized AtExitManager by itself, once it is done, we no longer need to call
InitAtExitManager in android_webview, even better, it might be done before
landing this patch 

+  g_at_exit_manager = new base::AtExitManager();
+}
+
 }  // namespace android
 }  // namespace base