Android's JVM requires threads to detach before they exit, but ONLY if
they needed to AttachCurrentThread. Conversly, threads that were
attached by the JVM (e.g. the result of making a native call from Java)
must NOT be detached by the application. This is bug 2441.
The fix for the above is to only pthread_setspecific() for threads that
Attach(), not for already-attached threads. To ensure that we only
detach Attached threads, added a GetEnv() call to ThreadDestructor(),
which revealed that Oracle's JVM can overly-eagerly clear TLS accounting
data, effectively detaching threads without their consent at shutdown.
Work around this with a specific check.
To guard against (some) regression, added a variant of PeerConnectionTest
that runs on a non-main thread. This revealed a bug in LinuxDeviceManager
which implicitly assumes its talk_base::Thread has already been
initialized. Fixed that here too.
BUG=2441
R=henrike@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/8759004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@5567 4adac7df-926f-26a2-2b94-8c16560cd09d
Also:
- Make PeerConnectionObserver::OnRenegotiationNeeded() pure virtual to avoid
this sort of mistake in the future.
- Sprinkle @Override annotations on some callback definitions that were missing
them.
- Fix a JNI method-signature-lookup typo (s/(V)V/()V/) in PCOJava::OnError()
- Add an explicit ScopedLocalFrameRef to PCOJava::OnError() to match all other
C++-fired callbacks, for consistency.
BUG=2771
R=wu@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/6829004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@5376 4adac7df-926f-26a2-2b94-8c16560cd09d
In r4665 I went a bit crazy with the manual reinterpretation of a pointer to a
jlong (to avoid undefined behavior) but that's what reinterpret_cast<> is for.
So use it directly now.
Added a do-nothing DataChannel to AppRTCDemo to regression test this, since the
only repro I've found of the original bug requires ARM ABI (PeerConnectionTest
on ia32 fails to repro).
BUG=2302
R=henrike@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/5489004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@5269 4adac7df-926f-26a2-2b94-8c16560cd09d
Analogous to the recent libjingle change: http://cl/54929753-p10.
This supports scoped_ptr<T[]> and scoped_ptr<C, FreeDeleter> rather
than scoped_array and scoped_ptr_malloc respectively.
- Add Chromium's template-based COMPILE_ASSERT. We didn't have this
previously in order to support the macro in C. Instead, move the
existing macro to compile_assert_c.h.
- Additionally copy the move.h and template_util.h depedencies and add
the WARN_UNUSED_RESULT macro.
- Leave scoped_array and scoped_ptr_malloc for now, but mark as
deprecated.
- Remove scoped_ptr foo(NULL) use. The default constructor handles it.
- Remove the now redundant COMPILE_ASSERT from peerconnection_jni.cc.
- Add a CHECK_ARRAY_SIZE macro to rtp_format_vp8_unittest.cc to remove
some repeated code.
TESTED=trybots
R=pbos@webrtc.org, tommi@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/2449005
git-svn-id: http://webrtc.googlecode.com/svn/trunk@5015 4adac7df-926f-26a2-2b94-8c16560cd09d
I mistakenly ommitted the checks when logging.h was ported from
libjingle to webrtc. This caused a significant CPU cost for logs which
were later filtered out anyway.
Verified with LS_VERBOSE logging in neteq4, running:
$ out/Release/modules_unittests \
--gtest_filter=NetEqDecodingTest.TestBitExactness \
--gtest_repeat=50 > time.txt
$ grep "case ran" time.txt | grep "[0-9]* ms" -o | sort
Results on a MacBook Retina, averaged over 5 runs:
Verbose logs disabled: 666 ms
Exisiting implementation, verbose logs enabled: 944 ms (1.42x)
New implementation, verbose logs enabled: 673 ms (1.01x)
BUG=2314
R=henrik.lundin@webrtc.org, henrike@webrtc.org, kjellander@webrtc.org, turaj@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/2160005
git-svn-id: http://webrtc.googlecode.com/svn/trunk@4682 4adac7df-926f-26a2-2b94-8c16560cd09d
- TCPPort::~TCPPort() was leaking incoming_ sockets; now they are deleted.
- PeerConnection::RemoveStream() now removes streams even if the
PeerConnection::IsClosed(). Previously such streams would never get removed.
- Gave MediaStreamTrackInterface a virtual destructor to ensure deletes on base
pointers are dispatched virtually.
- VideoTrack.dispose() delegates to super.dispose() (instead of leaking)
- PeerConnection.dispose() now removes streams before disposing of them.
- MediaStream.dispose() now removes tracks before disposing of them.
- VideoCapturer.dispose() only unowned VideoCapturers (mirroring C++ API)
- AppRTCDemo.disconnectAndExit() now correctly .dispose()s its
VideoSource and PeerConnectionFactory.
- CHECK that Release()d objects are deleted when expected to (i.e. no ref-cycles
or missing .dispose() calls) in the Java API.
- Create & Return webrtc::Traces at factory birth/death to be able to assert
that _all_ threads started during the test are collected by the end.
- Name threads attached to the JVM more informatively for debugging.
- Removed a bunch of unnecessary scoped_refptr instances in
peerconnection_jni.cc whose only job was messing with refcounts.
RISK=P2
TESTED=AppRTCDemo can be ended and restarted just fine instead of crashing on camera unavailability. No more post-app-exit logcat lines. PCTest.java now asserts that all threads are collected before exit.
BUG=2183
R=wu@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/2005004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@4534 4adac7df-926f-26a2-2b94-8c16560cd09d
