[Android] Fix a few issues related to old lib deletion.

base/android/java/src/org/chromium/base/library_loader/LibraryLoader.java

Index: base/android/java/src/org/chromium/base/library_loader/LibraryLoader.java
diff --git a/base/android/java/src/org/chromium/base/library_loader/LibraryLoader.java b/base/android/java/src/org/chromium/base/library_loader/LibraryLoader.java
index c3b3b35fc5ec0b8b6e4e9b5f868d88a5ddd8901b..22513810ec9cf90dd3f45453f3b604dc7c355779 100644
--- a/base/android/java/src/org/chromium/base/library_loader/LibraryLoader.java
+++ b/base/android/java/src/org/chromium/base/library_loader/LibraryLoader.java
@@ -47,14 +47,20 @@ public class LibraryLoader {
     // The flag is used to report UMA stats later.
     private static boolean sNativeLibraryHackWasUsed = false;
 
+    // TODO(fqian): Remove this method once downstream CL is
+    // committed.
+    public static void ensureInitialized(Context context) throws ProcessInitException {
+        ensureInitialized(context, true);
+    }
 
     /**
-     * TODO: http://crbug.com/354655
-     * remove this method once WebViewChromiumFactoryProvider.java
-     * changes the call to ensureInitialized(null).
+     * The same as ensureInitialized(null, false), should only be called
+     * by non-browser processes.
+     *
+     * @throws ProcessInitException
      */
     public static void ensureInitialized() throws ProcessInitException {
-        ensureInitialized(null);
+        ensureInitialized(null, false);
     }
 
     /**
@@ -70,14 +76,18 @@ public class LibraryLoader {
      *    will extract the native libraries from APK. This is a hack used to
      *    work around some Sony devices with the following platform bug:
      *    http://b/13216167.
+     *
+     *  @param inBrowserProcess The caller indicates it is running in the browser
+     *    process, so the code will perform browser-specific file cleanup.
      */
-    public static void ensureInitialized(Context context) throws ProcessInitException {
+    public static void ensureInitialized(Context context, boolean inBrowserProcess)
+            throws ProcessInitException {
         synchronized (sLock) {
             if (sInitialized) {
                 // Already initialized, nothing to do.
                 return;
             }
-            loadAlreadyLocked(context);
+            loadAlreadyLocked(context, inBrowserProcess);
             initializeAlreadyLocked(CommandLine.getJavaSwitchesOrNull());
         }
     }
@@ -92,17 +102,32 @@ public class LibraryLoader {
     }
 
     /**
+     * The same as loadNow(null, false), should only be called by
+     * non-browser process.
+     *
+     * @throws ProcessInitException
+     */
+    public static void loadNow() throws ProcessInitException {
+        loadNow(null, false);
+    }
+
+    /**
      * Loads the library and blocks until the load completes. The caller is responsible
      * for subsequently calling ensureInitialized().
      * May be called on any thread, but should only be called once. Note the thread
      * this is called on will be the thread that runs the native code's static initializers.
      * See the comment in doInBackground() for more considerations on this.
      *
+     * @param context The context the code is running, or null if it doesn't have one.
+     * @param inBrowserProcess The flag tells whether the code is running in
+     *   the browser process.

[cjhopman, 2014/04/02 22:24:24]

It's unclear (from this documentation) what this flag does. How about renaming
it to something like shouldDeleteOldWorkaroundLibraries (or something shorter).

Then it also won't feel so much like this "knows about" the browser process.

[Feng Qian, 2014/04/03 00:50:42]

ok, I am fine with renaming.

On 2014/04/02 22:24:24, cjhopman wrote:

+     *
      * @throws ProcessInitException if the native library failed to load.
      */
-    public static void loadNow(Context context) throws ProcessInitException {
+    public static void loadNow(Context context, boolean inBrowserProcess)
+            throws ProcessInitException {
         synchronized (sLock) {
-            loadAlreadyLocked(context);
+            loadAlreadyLocked(context, inBrowserProcess);
         }
     }
 
@@ -120,7 +145,8 @@ public class LibraryLoader {
     }
 
     // Invoke System.loadLibrary(...), triggering JNI_OnLoad in native code
-    private static void loadAlreadyLocked(Context context) throws ProcessInitException {
+    private static void loadAlreadyLocked(Context context, boolean inBrowserProcess)
+            throws ProcessInitException {
         try {
             if (!sLoaded) {
                 assert !sInitialized;
@@ -130,6 +156,7 @@ public class LibraryLoader {
 
                 if (useChromiumLinker) Linker.prepareLibraryLoad();
 
+                boolean workaroundLibrariesDeleted = false;
                 for (String library : NativeLibraries.LIBRARIES) {
                     Log.i(TAG, "Loading: " + library);
                     if (useChromiumLinker) {
@@ -137,9 +164,12 @@ public class LibraryLoader {
                     } else {
                         try {
                             System.loadLibrary(library);
-                            if (context != null) {
+                            if (context != null
+                                && inBrowserProcess

[boliu, 2014/04/03 01:03:22]

Webview technically only has the browser process, so "inBrowserProcess" is
misleading. Can we rename this to what it actually does, something like
"skipDeleteWorkaroundLibrary"?

Or just set it to true in webview, since context is NULL in webview anyway...

+                                && !workaroundLibrariesDeleted) {
                                 LibraryLoaderHelper.deleteWorkaroundLibrariesAsynchronously(
                                     context);
+                                workaroundLibrariesDeleted = true;

[cjhopman, 2014/04/02 22:24:24]

Since we only want this to happen once, how about moving it outside of the loop
over libraries. sNativeLibraryHackWasUsed is probably a better signal for
whether or not we should delete the workaround libraries anyway.

[Feng Qian, 2014/04/03 00:50:42]

The code is not just run when sNativeLibraryHackWasUsed is true.
But you bring up a good point that, even if the content linker is used for this
build, the previous build might used the hack, so the code should always run.

On 2014/04/02 22:24:24, cjhopman wrote:

                             }
                         } catch (UnsatisfiedLinkError e) {
                             if (context != null