[Android WebView] Close InputStreams when we're done with them.

android_webview/native/input_stream_impl.cc

Index: android_webview/native/input_stream_impl.cc
diff --git a/android_webview/native/input_stream_impl.cc b/android_webview/native/input_stream_impl.cc
index 5838650e43b5a74c2b51d8e2cb099317f206b934..5272ee54c1852bede7ba500fe7a240b2903d300a 100644
--- a/android_webview/native/input_stream_impl.cc
+++ b/android_webview/native/input_stream_impl.cc
@@ -18,6 +18,7 @@ using base::android::AttachCurrentThread;
 using base::android::ClearException;
 using base::android::JavaRef;
 using JNI_InputStream::Java_InputStream_available;
+using JNI_InputStream::Java_InputStream_close;
 using JNI_InputStream::Java_InputStream_skip;
 using JNI_InputStream::Java_InputStream_readI_AB_I_I;
 
@@ -45,6 +46,9 @@ InputStreamImpl::InputStreamImpl(const JavaRef<jobject>& stream)
 }
 
 InputStreamImpl::~InputStreamImpl() {
+  JNIEnv* env = AttachCurrentThread();
+  Java_InputStream_close(env, jobject_.obj());
+  DCHECK(!ClearException(env));

[joth, 2013/01/03 19:18:18]

the ClearException will be removed in release build, so it's better to just call
CheckException() unconditionally.

[mkosiba (inactive), 2013/01/04 12:31:11]

uh, actually due to the way the JNI methods are auto-generated the
Java_InputStream_close will CHECK(false) if we get an exception back.

The bug is crbug.com/157880.

[benm (inactive), 2013/01/04 12:55:11]

I see. So right now in the exception case we'll never hit this; but when that
bug is fixed we would still want to call CheckException() here?

[mkosiba (inactive), 2013/01/04 14:41:01]

When the bug is fixed we'll have an unsafe version of the method
(Java_InputStream_unsafe_close ?) which we would have to use instead in order to
have the ClearException code be reachable.
I'd suggest to file a bug to track this.

[benm (inactive), 2013/01/04 16:59:40]

Gotcha. Follow up in https://codereview.chromium.org/11748032/.

 }
 
 bool InputStreamImpl::BytesAvailable(int* bytes_available) const {