206068 2020-01-10 08:45:32 -0800 REGRESSION (r254291): [ Catalina wk2 Debug ] Flaky ASSERT on fast/images/animated-image-loop-count.html 2020-01-16 13:02:50 -0800 1 1 1 Unclassified WebKit Images WebKit Nightly Build Unspecified Unspecified RESOLVED FIXED https://bugs.webkit.org/show_bug.cgi?id=205850 https://bugs.webkit.org/show_bug.cgi?id=200734 https://bugs.webkit.org/show_bug.cgi?id=206326 InRadar P2 Normal --- 1 tsavell clord ap cdumez clord commit-queue dino eric.carlson sabouhallawa simon.fraser webkit-bot-watchers-bugzilla webkit-bug-importer oldest_to_newest 1605118 0 tsavell 2020-01-10 08:45:32 -0800 fast/images/animated-image-loop-count.html Description: This test seems to have begun crashing around r254298 only on Catalina wk2 Debug. this is not reproducing locally for m right now History: https://results.webkit.org/?suite=layout-tests&test=fast%2Fimages%2Fanimated-image-loop-count.html Crash Log: No crash log found for com.apple.WebKit.WebContent.Development:3398. stdout: stderr: 2 0x61b6f22eb WTFCrashWithInfo(int, char const*, char const*, int) 3 0x61efbfb66 WebCore::ImageSource::~ImageSource() 4 0x61efbfcf5 WebCore::ImageSource::~ImageSource() 5 0x61ef213ea WTF::ThreadSafeRefCounted<WebCore::ImageSource, (WTF::DestructionThread)0>::deref() const::'lambda'()::operator()() const 6 0x61ef213ad WTF::ThreadSafeRefCounted<WebCore::ImageSource, (WTF::DestructionThread)0>::deref() const 7 0x61ef2135f WTF::Ref<WebCore::ImageSource, WTF::DumbPtrTraits<WebCore::ImageSource> >::~Ref() 8 0x61eef23f5 WTF::Ref<WebCore::ImageSource, WTF::DumbPtrTraits<WebCore::ImageSource> >::~Ref() 9 0x61efc99ec WebCore::ImageSource::startAsyncDecodingQueue()::$_1::~$_1() 10 0x61efc1d85 WebCore::ImageSource::startAsyncDecodingQueue()::$_1::~$_1() 11 0x61efcfe31 WTF::Detail::CallableWrapper<WebCore::ImageSource::startAsyncDecodingQueue()::$_1, void>::~CallableWrapper() 12 0x61efcf915 WTF::Detail::CallableWrapper<WebCore::ImageSource::startAsyncDecodingQueue()::$_1, void>::~CallableWrapper() 13 0x61efcf939 WTF::Detail::CallableWrapper<WebCore::ImageSource::startAsyncDecodingQueue()::$_1, void>::~CallableWrapper() 14 0x635b2e03f std::__1::default_delete<WTF::Detail::CallableWrapperBase<void> >::operator()(WTF::Detail::CallableWrapperBase<void>*) const 15 0x635b2dfbf std::__1::unique_ptr<WTF::Detail::CallableWrapperBase<void>, std::__1::default_delete<WTF::Detail::CallableWrapperBase<void> > >::reset(WTF::Detail::CallableWrapperBase<void>*) 16 0x635b2df59 std::__1::unique_ptr<WTF::Detail::CallableWrapperBase<void>, std::__1::default_delete<WTF::Detail::CallableWrapperBase<void> > >::~unique_ptr() 17 0x635b2df35 std::__1::unique_ptr<WTF::Detail::CallableWrapperBase<void>, std::__1::default_delete<WTF::Detail::CallableWrapperBase<void> > >::~unique_ptr() 18 0x635b2df15 WTF::Function<void ()>::~Function() 19 0x635b2d215 WTF::Function<void ()>::~Function() 20 0x635c46de3 WTF::WorkQueue::dispatch(WTF::Function<void ()>&&)::$_0::~$_0() 21 0x635c46125 WTF::WorkQueue::dispatch(WTF::Function<void ()>&&)::$_0::~$_0() 22 0x635c467f0 WTF::BlockPtr<void ()> WTF::BlockPtr<void ()>::fromCallable<WTF::WorkQueue::dispatch(WTF::Function<void ()>&&)::$_0>(WTF::WorkQueue::dispatch(WTF::Function<void ()>&&)::$_0)::'lambda'(void const*)::operator()(void const*) const 23 0x635c46775 WTF::BlockPtr<void ()> WTF::BlockPtr<void ()>::fromCallable<WTF::WorkQueue::dispatch(WTF::Function<void ()>&&)::$_0>(WTF::WorkQueue::dispatch(WTF::Function<void ()>&&)::$_0)::'lambda'(void const*)::__invoke(void const*) 24 0x7fff709a0bed _Block_release 25 0x7fff708f94de _dispatch_client_callout 26 0x7fff708fea9e _dispatch_lane_serial_drain 27 0x7fff708ff422 _dispatch_lane_invoke 28 0x7fff70908aa1 _dispatch_workloop_worker_thread 29 0x7fff70b52763 _pthread_wqthread 30 0x7fff70b525c3 start_wqthread LEAK: 1 WebPageProxy 1605119 1 webkit-bug-importer 2020-01-10 08:45:52 -0800 <rdar://problem/58480028> 1605595 2 ap 2020-01-11 18:05:14 -0800 Seems almost certain to be https://trac.webkit.org/r254291 1605768 3 clord 2020-01-13 01:46:01 -0800 (In reply to Alexey Proskuryakov from comment #2) > Seems almost certain to be https://trac.webkit.org/r254291 Almost certainly, looking into it. 1605770 4 clord 2020-01-13 02:09:04 -0800 So I'm pretty certain this is caused by the async decoding queue causing destruction to happen off the creation thread (and then hitting the assert in the destructor) - this worked before because ImageSource was constructed with the main thread set as the destruction thread. I'm just testing a patch that moves the references back to the creation thread when exiting the async decoding queue task. I don't know if there's a more elegant way of fixing this(?) 1605784 5 387509 clord 2020-01-13 02:52:10 -0800 Created attachment 387509 Patch 1606051 6 387509 cdumez 2020-01-13 14:19:32 -0800 Comment on attachment 387509 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=387509&action=review I think this would work fine but I have a couple of nits. > Source/WebCore/platform/graphics/ImageSource.cpp:349 > + decodingQueue().dispatch([protectedThis = makeRef(*this), protectedDecodingQueue = makeRef(decodingQueue()), protectedFrameRequestQueue = makeRef(frameRequestQueue()), protectedDecoder = makeRef(*m_decoder), sourceURL = sourceURL().string().isolatedCopy()] () mutable { I do not understand why we're capturing the protectedDecodingQueue here, is does not appear to be needed. WorkQueue::dispatch() already takes care of ref'ing the WorkQueue object internally. > Source/WebCore/platform/graphics/ImageSource.cpp:385 > + // Ensure destruction happens on creation thread Missing period at the end of the comment (per coding style). > Source/WebCore/platform/graphics/ImageSource.cpp:386 > + protectedThis->m_runLoop.dispatch([protectedThis = WTFMove(protectedThis), protectedQueue = WTFMove(protectedDecodingQueue), protectedDecoder = WTFMove(protectedDecoder)] () mutable { }); I see no good reason to dispatch the protectedQueue here, I don't think it should even have been captured in the first place. 1606085 7 387509 sabouhallawa 2020-01-13 15:25:41 -0800 Comment on attachment 387509 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=387509&action=review >> Source/WebCore/platform/graphics/ImageSource.cpp:386 >> + protectedThis->m_runLoop.dispatch([protectedThis = WTFMove(protectedThis), protectedQueue = WTFMove(protectedDecodingQueue), protectedDecoder = WTFMove(protectedDecoder)] () mutable { }); > > I see no good reason to dispatch the protectedQueue here, I don't think it should even have been captured in the first place. Why did we even change this function in r254291? The async image decoding is disabled for 2D canvas and I believe it has to be disabled for off-screen canvas. So I think this function should be retuned it its state before r254291. Maybe we can assert at its beginning ASSERT(isMainThread()) since the async decoding is only enabled for HTMLImageElement and CSS background images. 1606306 8 387633 clord 2020-01-14 01:37:55 -0800 Created attachment 387633 Patch 1606308 9 clord 2020-01-14 01:42:25 -0800 Addressing the comments, I've changed both dispatches back to callOnMainThread, added a main-thread assert in the method and corrected the comment's missing full stop. I didn't change the protectedQueue ref - at some point a reference is required to the queue being used to make sure it's the same one when updating cached frames and that seems a fine way to do it. This function should now be the same as it was pre r254291 now, but with an added assert and explicitly moving the references back to the main thread when exiting the queue (as the object no longer has the main thread set as the destruction thread). 1606532 10 387633 sabouhallawa 2020-01-14 12:33:40 -0800 Comment on attachment 387633 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=387633&action=review > Source/WebCore/platform/graphics/ImageSource.cpp:389 > + > + // Ensure destruction happens on creation thread. > + callOnMainThread([protectedThis = WTFMove(protectedThis), protectedQueue = WTFMove(protectedDecodingQueue), protectedDecoder = WTFMove(protectedDecoder)] () mutable { }); Is this part still needed? I think r248703 ensures the destruction happens on the main thread. 1606671 11 clord 2020-01-14 16:19:32 -0800 (In reply to Said Abou-Hallawa from comment #10) > Comment on attachment 387633 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=387633&action=review > > > Source/WebCore/platform/graphics/ImageSource.cpp:389 > > + > > + // Ensure destruction happens on creation thread. > > + callOnMainThread([protectedThis = WTFMove(protectedThis), protectedQueue = WTFMove(protectedDecodingQueue), protectedDecoder = WTFMove(protectedDecoder)] () mutable { }); > > Is this part still needed? I think r248703 ensures the destruction happens > on the main thread. This is needed because r254291 essentially rolls back that change so that ImageSource is usable in workers (needed for ImageBitmap). 1607005 12 387633 cdumez 2020-01-15 10:54:06 -0800 Comment on attachment 387633 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=387633&action=review > Source/WebCore/platform/graphics/ImageSource.cpp:377 > + callOnMainThread([protectedThis = protectedThis.copyRef(), protectedQueue = protectedDecodingQueue.copyRef(), protectedDecoder = protectedDecoder.copyRef(), sourceURL = sourceURL.isolatedCopy(), nativeImage = WTFMove(nativeImage), frameRequest] () mutable { Please use WTFMove() here instead of copyRef()... >>> Source/WebCore/platform/graphics/ImageSource.cpp:389 >>> + callOnMainThread([protectedThis = WTFMove(protectedThis), protectedQueue = WTFMove(protectedDecodingQueue), protectedDecoder = WTFMove(protectedDecoder)] () mutable { }); >> >> Is this part still needed? I think r248703 ensures the destruction happens on the main thread. > > This is needed because r254291 essentially rolls back that change so that ImageSource is usable in workers (needed for ImageBitmap). ...and then you no longer need this callOnMainThread(). Make sense? 1607376 13 clord 2020-01-16 00:53:54 -0800 (In reply to Chris Dumez from comment #12) > Comment on attachment 387633 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=387633&action=review > > > Source/WebCore/platform/graphics/ImageSource.cpp:377 > > + callOnMainThread([protectedThis = protectedThis.copyRef(), protectedQueue = protectedDecodingQueue.copyRef(), protectedDecoder = protectedDecoder.copyRef(), sourceURL = sourceURL.isolatedCopy(), nativeImage = WTFMove(nativeImage), frameRequest] () mutable { > > Please use WTFMove() here instead of copyRef()... This call is in a loop that uses those objects, if we used WTFMove here, surely we'd just crash when the next frame is requested? (note, using copyRef here is not something I introduced, it was always like this) > >>> Source/WebCore/platform/graphics/ImageSource.cpp:389 > >>> + callOnMainThread([protectedThis = WTFMove(protectedThis), protectedQueue = WTFMove(protectedDecodingQueue), protectedDecoder = WTFMove(protectedDecoder)] () mutable { }); > >> > >> Is this part still needed? I think r248703 ensures the destruction happens on the main thread. > > > > This is needed because r254291 essentially rolls back that change so that ImageSource is usable in workers (needed for ImageBitmap). > > ...and then you no longer need this callOnMainThread(). > > Make sense? I don't think this makes sense, if I've missed something please elaborate. 1607522 14 387633 cdumez 2020-01-16 10:10:47 -0800 Comment on attachment 387633 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=387633&action=review >>>>> Source/WebCore/platform/graphics/ImageSource.cpp:389 >>>>> + callOnMainThread([protectedThis = WTFMove(protectedThis), protectedQueue = WTFMove(protectedDecodingQueue), protectedDecoder = WTFMove(protectedDecoder)] () mutable { }); >>>> >>>> Is this part still needed? I think r248703 ensures the destruction happens on the main thread. >>> >>> This is needed because r254291 essentially rolls back that change so that ImageSource is usable in workers (needed for ImageBitmap). >> >> ...and then you no longer need this callOnMainThread(). >> >> Make sense? > > I don't think this makes sense, if I've missed something please elaborate. Now that I expand the diff, I see that you are in a while loop and therefore cannot use WTFMove(above), I thought we were in the same scope as your callOnMainThread() here. Never mind my comment then, I do agree that this new callOnMainThread is needed. 1607532 15 387633 commit-queue 2020-01-16 10:30:20 -0800 Comment on attachment 387633 Patch Clearing flags on attachment: 387633 Committed r254692: <https://trac.webkit.org/changeset/254692> 1607533 16 commit-queue 2020-01-16 10:30:22 -0800 All reviewed patches have been landed. Closing bug. 1607612 17 ap 2020-01-16 13:02:50 -0800 *** Bug 206326 has been marked as a duplicate of this bug. *** 387509 2020-01-13 02:52:10 -0800 2020-01-14 01:37:52 -0800 Patch bug-206068-20200113105209.patch text/plain 2579 clord U3VidmVyc2lvbiBSZXZpc2lvbjogMjU0NDIyCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggMGRhNjVjNjY5MWY0ZmQw ZDAzOGE0MzYzMDU3NzMxMjllZDdmNGFlNC4uNjc2M2E5NTgyNGZlNzJmYWUxMGZkN2ZmYjZiYmNm ZmU1OTM2MzczMSAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1Nv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDE2IEBACisyMDIwLTAxLTEzICBDaHJp cyBMb3JkICA8Y2xvcmRAaWdhbGlhLmNvbT4KKworICAgICAgICBSRUdSRVNTSU9OIChyMjU0Mjkx KTogWyBDYXRhbGluYSB3azIgRGVidWcgXSBGbGFreSBBU1NFUlQgb24gZmFzdC9pbWFnZXMvYW5p bWF0ZWQtaW1hZ2UtbG9vcC1jb3VudC5odG1sCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQu b3JnL3Nob3dfYnVnLmNnaT9pZD0yMDYwNjgKKyAgICAgICAgPHJkYXI6Ly9wcm9ibGVtLzU4NDgw MDI4PgorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIE5v IG5ldyB0ZXN0cywgY292ZXJlZCBieSBleGlzdGluZyB0ZXN0cy4KKworICAgICAgICAqIHBsYXRm b3JtL2dyYXBoaWNzL0ltYWdlU291cmNlLmNwcDoKKyAgICAgICAgKFdlYkNvcmU6OkltYWdlU291 cmNlOjpzdGFydEFzeW5jRGVjb2RpbmdRdWV1ZSk6CisKIDIwMjAtMDEtMTIgIFphbGFuIEJ1anRh cyAgPHphbGFuQGFwcGxlLmNvbT4KIAogICAgICAgICBUZXh0LWluZGVudCB3aXRoIHBlcmNlbnRh Z2UgdmFsdWUgc2hvdWxkIHJlc29sdmUgYWdhaW5zdCB0aGUgYXZhaWxhYmxlIHdpZHRoCmRpZmYg LS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9JbWFnZVNvdXJjZS5jcHAg Yi9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9JbWFnZVNvdXJjZS5jcHAKaW5kZXgg NWVhYTczMmZlZmY4ZmFlYjA5MTllY2VmMWY3M2FiMDJiMDU0YWU5MC4uNTE4MzY5ZWE4Zjc0OTI5 N2Y5NWVkM2RkNzIzOGQ3NWQyMGQ4Yjg5MyAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvcGxh dGZvcm0vZ3JhcGhpY3MvSW1hZ2VTb3VyY2UuY3BwCisrKyBiL1NvdXJjZS9XZWJDb3JlL3BsYXRm b3JtL2dyYXBoaWNzL0ltYWdlU291cmNlLmNwcApAQCAtMzQ2LDcgKzM0Niw3IEBAIHZvaWQgSW1h Z2VTb3VyY2U6OnN0YXJ0QXN5bmNEZWNvZGluZ1F1ZXVlKCkKICAgICAgICAgcmV0dXJuOwogCiAg ICAgLy8gV2UgbmVlZCB0byBwcm90ZWN0IHRoaXMsIG1fZGVjb2RpbmdRdWV1ZSBhbmQgbV9kZWNv ZGVyIGZyb20gYmVpbmcgZGVsZXRlZCB3aGlsZSB3ZSBhcmUgaW4gdGhlIGRlY29kaW5nIGxvb3Au Ci0gICAgZGVjb2RpbmdRdWV1ZSgpLmRpc3BhdGNoKFtwcm90ZWN0ZWRUaGlzID0gbWFrZVJlZigq dGhpcyksIHByb3RlY3RlZERlY29kaW5nUXVldWUgPSBtYWtlUmVmKGRlY29kaW5nUXVldWUoKSks IHByb3RlY3RlZEZyYW1lUmVxdWVzdFF1ZXVlID0gbWFrZVJlZihmcmFtZVJlcXVlc3RRdWV1ZSgp KSwgcHJvdGVjdGVkRGVjb2RlciA9IG1ha2VSZWYoKm1fZGVjb2RlciksIHNvdXJjZVVSTCA9IHNv dXJjZVVSTCgpLnN0cmluZygpLmlzb2xhdGVkQ29weSgpXSB7CisgICAgZGVjb2RpbmdRdWV1ZSgp LmRpc3BhdGNoKFtwcm90ZWN0ZWRUaGlzID0gbWFrZVJlZigqdGhpcyksIHByb3RlY3RlZERlY29k aW5nUXVldWUgPSBtYWtlUmVmKGRlY29kaW5nUXVldWUoKSksIHByb3RlY3RlZEZyYW1lUmVxdWVz dFF1ZXVlID0gbWFrZVJlZihmcmFtZVJlcXVlc3RRdWV1ZSgpKSwgcHJvdGVjdGVkRGVjb2RlciA9 IG1ha2VSZWYoKm1fZGVjb2RlciksIHNvdXJjZVVSTCA9IHNvdXJjZVVSTCgpLnN0cmluZygpLmlz b2xhdGVkQ29weSgpXSAoKSBtdXRhYmxlIHsKICAgICAgICAgSW1hZ2VGcmFtZVJlcXVlc3QgZnJh bWVSZXF1ZXN0OwogICAgICAgICBTZWNvbmRzIG1pbkRlY29kaW5nRHVyYXRpb24gPSBwcm90ZWN0 ZWRUaGlzLT5mcmFtZURlY29kaW5nRHVyYXRpb25Gb3JUZXN0aW5nKCk7CiAKQEAgLTM4MSw2ICsz ODEsOSBAQCB2b2lkIEltYWdlU291cmNlOjpzdGFydEFzeW5jRGVjb2RpbmdRdWV1ZSgpCiAgICAg ICAgICAgICAgICAgICAgIExPRyhJbWFnZXMsICJJbWFnZVNvdXJjZTo6JXMgLSAlcCAtIHVybDog JXMgW2ZyYW1lICVsZCB3aWxsIG5vdCBjYWNoZWRdIiwgX19GVU5DVElPTl9fLCBwcm90ZWN0ZWRU aGlzLnB0cigpLCBzb3VyY2VVUkwudXRmOCgpLmRhdGEoKSwgZnJhbWVSZXF1ZXN0LmluZGV4KTsK ICAgICAgICAgICAgIH0pOwogICAgICAgICB9CisKKyAgICAgICAgLy8gRW5zdXJlIGRlc3RydWN0 aW9uIGhhcHBlbnMgb24gY3JlYXRpb24gdGhyZWFkCisgICAgICAgIHByb3RlY3RlZFRoaXMtPm1f cnVuTG9vcC5kaXNwYXRjaChbcHJvdGVjdGVkVGhpcyA9IFdURk1vdmUocHJvdGVjdGVkVGhpcyks IHByb3RlY3RlZFF1ZXVlID0gV1RGTW92ZShwcm90ZWN0ZWREZWNvZGluZ1F1ZXVlKSwgcHJvdGVj dGVkRGVjb2RlciA9IFdURk1vdmUocHJvdGVjdGVkRGVjb2RlcildICgpIG11dGFibGUgeyB9KTsK ICAgICB9KTsKIH0KIAo= 387633 2020-01-14 01:37:55 -0800 2020-01-16 10:30:20 -0800 Patch bug-206068-20200114093754.patch text/plain 3919 clord U3VidmVyc2lvbiBSZXZpc2lvbjogMjU0NDIyCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggMGRhNjVjNjY5MWY0ZmQw ZDAzOGE0MzYzMDU3NzMxMjllZDdmNGFlNC4uMmE5MTM5ZmJkMDVjMjlmYzc2OGJlOGUwZmIyN2Y2 ZmFlNzRmY2JjMyAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1Nv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDE2IEBACisyMDIwLTAxLTEzICBDaHJp cyBMb3JkICA8Y2xvcmRAaWdhbGlhLmNvbT4KKworICAgICAgICBSRUdSRVNTSU9OIChyMjU0Mjkx KTogWyBDYXRhbGluYSB3azIgRGVidWcgXSBGbGFreSBBU1NFUlQgb24gZmFzdC9pbWFnZXMvYW5p bWF0ZWQtaW1hZ2UtbG9vcC1jb3VudC5odG1sCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQu b3JnL3Nob3dfYnVnLmNnaT9pZD0yMDYwNjgKKyAgICAgICAgPHJkYXI6Ly9wcm9ibGVtLzU4NDgw MDI4PgorCisgICAgICAgIFJldmlld2VkIGJ5IENocmlzIER1bWV6LgorCisgICAgICAgIE5vIG5l dyB0ZXN0cywgY292ZXJlZCBieSBleGlzdGluZyB0ZXN0cy4KKworICAgICAgICAqIHBsYXRmb3Jt L2dyYXBoaWNzL0ltYWdlU291cmNlLmNwcDoKKyAgICAgICAgKFdlYkNvcmU6OkltYWdlU291cmNl OjpzdGFydEFzeW5jRGVjb2RpbmdRdWV1ZSk6CisKIDIwMjAtMDEtMTIgIFphbGFuIEJ1anRhcyAg PHphbGFuQGFwcGxlLmNvbT4KIAogICAgICAgICBUZXh0LWluZGVudCB3aXRoIHBlcmNlbnRhZ2Ug dmFsdWUgc2hvdWxkIHJlc29sdmUgYWdhaW5zdCB0aGUgYXZhaWxhYmxlIHdpZHRoCmRpZmYgLS1n aXQgYS9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9JbWFnZVNvdXJjZS5jcHAgYi9T b3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9JbWFnZVNvdXJjZS5jcHAKaW5kZXggNWVh YTczMmZlZmY4ZmFlYjA5MTllY2VmMWY3M2FiMDJiMDU0YWU5MC4uMjcyNGQ3MzhlNzVmM2FkOGE3 NWEyYmU0MTU1MTA1YzUwY2JmODZmYiAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvcGxhdGZv cm0vZ3JhcGhpY3MvSW1hZ2VTb3VyY2UuY3BwCisrKyBiL1NvdXJjZS9XZWJDb3JlL3BsYXRmb3Jt L2dyYXBoaWNzL0ltYWdlU291cmNlLmNwcApAQCAtMzQ1LDggKzM0NSwxMSBAQCB2b2lkIEltYWdl U291cmNlOjpzdGFydEFzeW5jRGVjb2RpbmdRdWV1ZSgpCiAgICAgaWYgKGhhc0FzeW5jRGVjb2Rp bmdRdWV1ZSgpIHx8ICFpc0RlY29kZXJBdmFpbGFibGUoKSkKICAgICAgICAgcmV0dXJuOwogCisg ICAgLy8gQXN5bmMgZGVjb2RpbmcgaXMgb25seSBlbmFibGVkIGZvciBIVE1MSW1hZ2VFbGVtZW50 IGFuZCBDU1MgYmFja2dyb3VuZCBpbWFnZXMuCisgICAgQVNTRVJUKGlzTWFpblRocmVhZCgpKTsK KwogICAgIC8vIFdlIG5lZWQgdG8gcHJvdGVjdCB0aGlzLCBtX2RlY29kaW5nUXVldWUgYW5kIG1f ZGVjb2RlciBmcm9tIGJlaW5nIGRlbGV0ZWQgd2hpbGUgd2UgYXJlIGluIHRoZSBkZWNvZGluZyBs b29wLgotICAgIGRlY29kaW5nUXVldWUoKS5kaXNwYXRjaChbcHJvdGVjdGVkVGhpcyA9IG1ha2VS ZWYoKnRoaXMpLCBwcm90ZWN0ZWREZWNvZGluZ1F1ZXVlID0gbWFrZVJlZihkZWNvZGluZ1F1ZXVl KCkpLCBwcm90ZWN0ZWRGcmFtZVJlcXVlc3RRdWV1ZSA9IG1ha2VSZWYoZnJhbWVSZXF1ZXN0UXVl dWUoKSksIHByb3RlY3RlZERlY29kZXIgPSBtYWtlUmVmKCptX2RlY29kZXIpLCBzb3VyY2VVUkwg PSBzb3VyY2VVUkwoKS5zdHJpbmcoKS5pc29sYXRlZENvcHkoKV0geworICAgIGRlY29kaW5nUXVl dWUoKS5kaXNwYXRjaChbcHJvdGVjdGVkVGhpcyA9IG1ha2VSZWYoKnRoaXMpLCBwcm90ZWN0ZWRE ZWNvZGluZ1F1ZXVlID0gbWFrZVJlZihkZWNvZGluZ1F1ZXVlKCkpLCBwcm90ZWN0ZWRGcmFtZVJl cXVlc3RRdWV1ZSA9IG1ha2VSZWYoZnJhbWVSZXF1ZXN0UXVldWUoKSksIHByb3RlY3RlZERlY29k ZXIgPSBtYWtlUmVmKCptX2RlY29kZXIpLCBzb3VyY2VVUkwgPSBzb3VyY2VVUkwoKS5zdHJpbmco KS5pc29sYXRlZENvcHkoKV0gKCkgbXV0YWJsZSB7CiAgICAgICAgIEltYWdlRnJhbWVSZXF1ZXN0 IGZyYW1lUmVxdWVzdDsKICAgICAgICAgU2Vjb25kcyBtaW5EZWNvZGluZ0R1cmF0aW9uID0gcHJv dGVjdGVkVGhpcy0+ZnJhbWVEZWNvZGluZ0R1cmF0aW9uRm9yVGVzdGluZygpOwogCkBAIC0zNzEs NyArMzc0LDcgQEAgdm9pZCBJbWFnZVNvdXJjZTo6c3RhcnRBc3luY0RlY29kaW5nUXVldWUoKQog ICAgICAgICAgICAgICAgIHNsZWVwKG1pbkRlY29kaW5nRHVyYXRpb24gLSAoTW9ub3RvbmljVGlt ZTo6bm93KCkgLSBzdGFydGluZ1RpbWUpKTsKIAogICAgICAgICAgICAgLy8gVXBkYXRlIHRoZSBj YWNoZWQgZnJhbWVzIG9uIHRoZSBjcmVhdGlvbiB0aHJlYWQgdG8gYXZvaWQgdXBkYXRpbmcgdGhl IE1lbW9yeUNhY2hlIGZyb20gYSBkaWZmZXJlbnQgdGhyZWFkLgotICAgICAgICAgICAgcHJvdGVj dGVkVGhpcy0+bV9ydW5Mb29wLmRpc3BhdGNoKFtwcm90ZWN0ZWRUaGlzID0gcHJvdGVjdGVkVGhp cy5jb3B5UmVmKCksIHByb3RlY3RlZFF1ZXVlID0gcHJvdGVjdGVkRGVjb2RpbmdRdWV1ZS5jb3B5 UmVmKCksIHByb3RlY3RlZERlY29kZXIgPSBwcm90ZWN0ZWREZWNvZGVyLmNvcHlSZWYoKSwgc291 cmNlVVJMID0gc291cmNlVVJMLmlzb2xhdGVkQ29weSgpLCBuYXRpdmVJbWFnZSA9IFdURk1vdmUo bmF0aXZlSW1hZ2UpLCBmcmFtZVJlcXVlc3RdICgpIG11dGFibGUgeworICAgICAgICAgICAgY2Fs bE9uTWFpblRocmVhZChbcHJvdGVjdGVkVGhpcyA9IHByb3RlY3RlZFRoaXMuY29weVJlZigpLCBw cm90ZWN0ZWRRdWV1ZSA9IHByb3RlY3RlZERlY29kaW5nUXVldWUuY29weVJlZigpLCBwcm90ZWN0 ZWREZWNvZGVyID0gcHJvdGVjdGVkRGVjb2Rlci5jb3B5UmVmKCksIHNvdXJjZVVSTCA9IHNvdXJj ZVVSTC5pc29sYXRlZENvcHkoKSwgbmF0aXZlSW1hZ2UgPSBXVEZNb3ZlKG5hdGl2ZUltYWdlKSwg ZnJhbWVSZXF1ZXN0XSAoKSBtdXRhYmxlIHsKICAgICAgICAgICAgICAgICAvLyBUaGUgcXVldWUg bWF5IGhhdmUgYmVlbiBjbG9zZWQgaWYgYWZ0ZXIgd2UgZ290IHRoZSBmcmFtZSBOYXRpdmVJbWFn ZSwgc3RvcEFzeW5jRGVjb2RpbmdRdWV1ZSgpIHdhcyBjYWxsZWQuCiAgICAgICAgICAgICAgICAg aWYgKHByb3RlY3RlZFF1ZXVlLnB0cigpID09IHByb3RlY3RlZFRoaXMtPm1fZGVjb2RpbmdRdWV1 ZSAmJiBwcm90ZWN0ZWREZWNvZGVyLnB0cigpID09IHByb3RlY3RlZFRoaXMtPm1fZGVjb2Rlcikg ewogICAgICAgICAgICAgICAgICAgICBBU1NFUlQocHJvdGVjdGVkVGhpcy0+bV9mcmFtZUNvbW1p dFF1ZXVlLmZpcnN0KCkgPT0gZnJhbWVSZXF1ZXN0KTsKQEAgLTM4MSw2ICszODQsOSBAQCB2b2lk IEltYWdlU291cmNlOjpzdGFydEFzeW5jRGVjb2RpbmdRdWV1ZSgpCiAgICAgICAgICAgICAgICAg ICAgIExPRyhJbWFnZXMsICJJbWFnZVNvdXJjZTo6JXMgLSAlcCAtIHVybDogJXMgW2ZyYW1lICVs ZCB3aWxsIG5vdCBjYWNoZWRdIiwgX19GVU5DVElPTl9fLCBwcm90ZWN0ZWRUaGlzLnB0cigpLCBz b3VyY2VVUkwudXRmOCgpLmRhdGEoKSwgZnJhbWVSZXF1ZXN0LmluZGV4KTsKICAgICAgICAgICAg IH0pOwogICAgICAgICB9CisKKyAgICAgICAgLy8gRW5zdXJlIGRlc3RydWN0aW9uIGhhcHBlbnMg b24gY3JlYXRpb24gdGhyZWFkLgorICAgICAgICBjYWxsT25NYWluVGhyZWFkKFtwcm90ZWN0ZWRU aGlzID0gV1RGTW92ZShwcm90ZWN0ZWRUaGlzKSwgcHJvdGVjdGVkUXVldWUgPSBXVEZNb3ZlKHBy b3RlY3RlZERlY29kaW5nUXVldWUpLCBwcm90ZWN0ZWREZWNvZGVyID0gV1RGTW92ZShwcm90ZWN0 ZWREZWNvZGVyKV0gKCkgbXV0YWJsZSB7IH0pOwogICAgIH0pOwogfQogCg==