67516 2011-09-02 11:48:25 -0700 [V8] JavaScriptAudioNode should not declare that it mixes in EventTarget 2013-09-12 22:31:14 -0700 1 1 1 Unclassified WebKit DOM 528+ (Nightly build) Unspecified Unspecified RESOLVED WONTFIX P2 Normal --- 67312 1 dominicc dominicc andersca arv crogers oldest_to_newest 461391 0 dominicc 2011-09-02 11:48:25 -0700 This is part of larger clean-up explained in meta bug 67312. JavaScriptAudioNode is #ifdef'd to declare it mixes in EventTarget in the V8 bindings, but we probably don’t want to do this because: 1. It is not an event target per the spec <https://dvcs.w3.org/hg/audio/raw-file/tip/webaudio/specification.html#JavaScriptAudioNode-section> 2. It doesn’t have addEventListener, removeEventListener and dispatchEvent, which event targets should have. Because JavaScriptAudioNode has a callback, it wants to opt in to V8’s event-listener-keepalive-backreference scheme. JSC splits this into a separate "mark children" concept. For now we need to special-case JavaScriptAudioNode in the V8 code generator, like we do for SVGElementInstance. 461396 1 106171 dominicc 2011-09-02 11:51:01 -0700 Created attachment 106171 Patch 461454 2 106180 dominicc 2011-09-02 13:04:57 -0700 Created attachment 106180 Patch 461537 3 106180 crogers 2011-09-02 14:47:07 -0700 Comment on attachment 106180 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=106180&action=review > Source/WebCore/bindings/js/JSEventTarget.cpp:-182 > - return toJS(exec, globalObject, jsAudioNode); I notice you're removing this from the JSC code-path. Do you also need to make an appropriate change in CodeGeneratorJS.pm? I agree we should move over to Callback, but I don't want to land a change which will break web audio on any of the ports 461595 4 dominicc 2011-09-02 15:49:30 -0700 (In reply to comment #3) > (From update of attachment 106180 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=106180&action=review > > > Source/WebCore/bindings/js/JSEventTarget.cpp:-182 > > - return toJS(exec, globalObject, jsAudioNode); > > I notice you're removing this from the JSC code-path. Do you also need to make an appropriate change in CodeGeneratorJS.pm? You’re right; this could be problematic for JSC. > I agree we should move over to Callback, but I don't want to land a change which will break web audio on any of the ports Is JSC WebAudio in a buildable state? I may as well make onaudioprocess a callback. It would be nice if JavaScriptAudioNode had unit tests. Can you help me write one? 461599 5 dominicc 2011-09-02 15:52:55 -0700 Two more follow-up questions: Do any other browsers have implementations/are working on implementations of Web Audio? Who is a good person to ask to review Web Audio patches? 461616 6 crogers 2011-09-02 16:29:41 -0700 (In reply to comment #4) > (In reply to comment #3) > > (From update of attachment 106180 [details] [details]) > > View in context: https://bugs.webkit.org/attachment.cgi?id=106180&action=review > > > > > Source/WebCore/bindings/js/JSEventTarget.cpp:-182 > > > - return toJS(exec, globalObject, jsAudioNode); > > > > I notice you're removing this from the JSC code-path. Do you also need to make an appropriate change in CodeGeneratorJS.pm? > > You’re right; this could be problematic for JSC. > > > I agree we should move over to Callback, but I don't want to land a change which will break web audio on any of the ports > > Is JSC WebAudio in a buildable state? I may as well make onaudioprocess a callback. Yes, it's possible to build for the mac port and the GTK project has an experimental version running, although their changes are not committed yet. I think it's best to have the complete change from EventListener -> Callback working before we land. I can help you test that this is working properly. You can also test this somewhat yourself by making sure that this page runs correctly both before and after the change: http://chromium.googlecode.com/svn/trunk/samples/audio/javascript-processing.html You can "view source" to see how this works -- pretty simple in this case (except for the WebGL part which isn't relevant to our test) > > It would be nice if JavaScriptAudioNode had unit tests. Can you help me write one? Yes, probably it would simply involve verifying that the callback is firing repeatedly. 461618 7 crogers 2011-09-02 16:30:32 -0700 (In reply to comment #5) > Two more follow-up questions: > > Do any other browsers have implementations/are working on implementations of Web Audio? > > Who is a good person to ask to review Web Audio patches? I'd like to be on the review list, and Kenneth Russell does many Web Audio reviews. 462229 8 dominicc 2011-09-05 12:47:43 -0700 Should the callback take four parameters: void processAudio(JavaScriptAudioNode node, float playbackTime, sequence<AudioBuffer> inputBuffer, sequence<AudioBuffer> outputBuffer) Or should it take one object with these as properties? If it takes an object, what should the type of that object be named? (AudioProcessEvent isn’t a good name, because the object is no longer an event.) 462237 9 crogers 2011-09-05 13:12:19 -0700 (In reply to comment #8) > Should the callback take four parameters: > > void processAudio(JavaScriptAudioNode node, float playbackTime, sequence<AudioBuffer> inputBuffer, sequence<AudioBuffer> outputBuffer) > > Or should it take one object with these as properties? If it takes an object, what should the type of that object be named? (AudioProcessEvent isn’t a good name, because the object is no longer an event.) I don't want to change the existing API and break code that's already out in the field, so let's stick with one object with these as properties. How about re-naming like this "AudioProcessingEvent" -> "AudioProcessingCallbackInfo" ? Please also note that the current implementation has both "inputBuffer" and "outputBuffer" passed directly as type "AudioBuffer" instead of "sequence<AudioBuffer>" (as the spec currently describes). I'll change the spec to match the current behavior along with the new name we choose to replace AudioProcessingEvent. 518108 10 dominicc 2011-12-08 21:51:10 -0800 So the proposed IDL is something like: interface JavaScriptAudioNode : AudioNode { attribute AudioProcessingCallback onaudioprocess; readonly attribute long bufferSize; } [Callback, NoInterfaceObject] interface AudioProcessingCallback { void processAudio(AudioProcessingCallbackInfo info); } [NoInterfaceObject] interface AudioProcessingCallbackInfo { attribute JavaScriptAudioNode node; readonly attribute float playbackTime; readonly attribute AudioBuffer inputBuffer; readonly attribute AudioBuffer outputBuffer; } 518980 11 crogers 2011-12-11 19:45:42 -0800 Dominic, yes this looks about right. I'm assuming that "processAudio()" is just internal IDL language for the C++ part and has no meaning in JavaScript? In other words, the actual JavaScript callback function can be any name. Sorry if I'm being stupid - just want to make sure. We can change "playbackTime" to be double instead of float -- see: https://bugs.webkit.org/show_bug.cgi?id=74030 519015 12 dominicc 2011-12-11 22:00:30 -0800 (In reply to comment #11) > Dominic, yes this looks about right. I'm assuming that "processAudio()" is just internal IDL language for the C++ part and has no meaning in JavaScript? In other words, the actual JavaScript callback function can be any name. Sorry if I'm being stupid - just want to make sure. It depends. If you do this: var n = /* some JavaScriptAudioNode */ n.onaudioprocess = f; then f can be a function with any name. However if you do n = onaudioprocess = o; where o is an object that is not a function, then it will try to call o.processAudio(…). If that should be a different name, we can change it. > We can change "playbackTime" to be double instead of float -- see: > https://bugs.webkit.org/show_bug.cgi?id=74030 Got it. Thanks. 519025 13 crogers 2011-12-11 22:54:35 -0800 (In reply to comment #12) > (In reply to comment #11) > > Dominic, yes this looks about right. I'm assuming that "processAudio()" is just internal IDL language for the C++ part and has no meaning in JavaScript? In other words, the actual JavaScript callback function can be any name. Sorry if I'm being stupid - just want to make sure. > > It depends. If you do this: > > var n = /* some JavaScriptAudioNode */ > n.onaudioprocess = f; > > then f can be a function with any name. However if you do > > n = onaudioprocess = o; > > where o is an object that is not a function, then it will try to call o.processAudio(…). If that should be a different name, we can change it. Oh right, now I remember you explaining this to me. That sounds fine. 928640 14 andersca 2013-09-12 22:31:14 -0700 V8 is gone. 106171 2011-09-02 11:51:01 -0700 2011-09-02 13:04:53 -0700 Patch bug-67516-20110902115100.patch text/plain 2276 dominicc U3VidmVyc2lvbiBSZXZpc2lvbjogOTQ0MzUKZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL0No YW5nZUxvZyBiL1NvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwppbmRleCA4MWZjMzQ5OWMwZGE3NjA4 ODIxMWNhMmVjMTc5ZWIwZWRkYTg4OWIyLi4yMzc0NTM1NDRlYWZhNmQyYmYwZDBmY2MzNWRkZDhm ZjEzYjkwZGFjIDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvU291 cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTYgQEAKKzIwMTEtMDktMDIgIERvbWlu aWMgQ29vbmV5ICA8ZG9taW5pY2NAY2hyb21pdW0ub3JnPgorCisgICAgICAgIE1ha2UgSmF2YVNj cmlwdEF1ZGlvTm9kZSBubyBsb25nZXIgY2xhaW0gdG8gYmUgYW4gRXZlbnRUYXJnZXQuIEl0IGlz IG5vdC4KKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTY3 NTE2CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAgVGVz dHM6IExheW91dFRlc3RzL3dlYmF1ZGlvLCBhbHRob3VnaCBKYXZhU2NyaXB0QXVkaW9Ob2RlIGlz IG5vdCB0ZXN0ZWQgOigKKworICAgICAgICAqIGJpbmRpbmdzL3NjcmlwdHMvQ29kZUdlbmVyYXRv clY4LnBtOgorICAgICAgICAoR2V0SW50ZXJuYWxGaWVsZHMpOiBKYXZhU2NyaXB0QXVkaW9Ob2Rl IHdhbnRzIHRvIGtlZXAgaXRzIG9uYXVkaW9wcm9jZXNzIGNhbGxiYWNrIGFsaXZlLgorICAgICAg ICAqIHdlYmF1ZGlvL0phdmFTY3JpcHRBdWRpb05vZGUuaWRsOiBEb24ndCBjbGFpbSB0byBiZSBF dmVudFRhcmdldC4KKwogMjAxMS0wOS0wMiAgU2FtIFdlaW5pZyAgPHNhbUB3ZWJraXQub3JnPgog CiAgICAgICAgIFJlbW92ZSBCZWZvcmVQcm9jZXNzRXZlbnQsIGl0IHdhcyBuZXZlciBtZWFudCB0 byBiZQpkaWZmIC0tZ2l0IGEvU291cmNlL1dlYkNvcmUvYmluZGluZ3Mvc2NyaXB0cy9Db2RlR2Vu ZXJhdG9yVjgucG0gYi9Tb3VyY2UvV2ViQ29yZS9iaW5kaW5ncy9zY3JpcHRzL0NvZGVHZW5lcmF0 b3JWOC5wbQppbmRleCA5ZTI2MzQwNWI3NDUyZGVmNGJlNTRhMzk5MGY3ZTFhOWEwZjZkOGUzLi4z YjVhYzFmYWJmYjUzMmM0ODc1ZGFhMzljMGY1NjEwYjg0YmJmZjhiIDEwMDY0NAotLS0gYS9Tb3Vy Y2UvV2ViQ29yZS9iaW5kaW5ncy9zY3JpcHRzL0NvZGVHZW5lcmF0b3JWOC5wbQorKysgYi9Tb3Vy Y2UvV2ViQ29yZS9iaW5kaW5ncy9zY3JpcHRzL0NvZGVHZW5lcmF0b3JWOC5wbQpAQCAtNTA4LDYg KzUwOCw3IEBAIHN1YiBHZXRJbnRlcm5hbEZpZWxkcwogICAgICAgICAoJGRhdGFOb2RlLT5leHRl bmRlZEF0dHJpYnV0ZXMtPnsiRXZlbnRUYXJnZXQifSB8fAogICAgICAgICAgSXNTdWJUeXBlKCRk YXRhTm9kZSwgIkFic3RyYWN0V29ya2VyIikgfHwKICAgICAgICAgIElzU3ViVHlwZSgkZGF0YU5v ZGUsICJXb3JrZXJDb250ZXh0IikgfHwKKyAgICAgICAgICRuYW1lIGVxICJKYXZhU2NyaXB0QXVk aW9Ob2RlIiB8fAogICAgICAgICAgJG5hbWUgZXEgIlNWR0VsZW1lbnRJbnN0YW5jZSIpKSB7CiAg ICAgICAgIHB1c2goQGN1c3RvbUludGVybmFsRmllbGRzLCAiZXZlbnRMaXN0ZW5lckNhY2hlSW5k ZXgiKTsKICAgICB9CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS93ZWJhdWRpby9KYXZhU2Ny aXB0QXVkaW9Ob2RlLmlkbCBiL1NvdXJjZS9XZWJDb3JlL3dlYmF1ZGlvL0phdmFTY3JpcHRBdWRp b05vZGUuaWRsCmluZGV4IGVmNTM1OWI5NzZlYWM2MzIxZDE4NWM5MWZhNjk1MjE1ZWZlNzg5NTku LmIxMDI4NDk5ZDQ0ODIyYjA5ZmEwY2RjZTM2MTFkOTZiOGE5NTQxZGEgMTAwNjQ0Ci0tLSBhL1Nv dXJjZS9XZWJDb3JlL3dlYmF1ZGlvL0phdmFTY3JpcHRBdWRpb05vZGUuaWRsCisrKyBiL1NvdXJj ZS9XZWJDb3JlL3dlYmF1ZGlvL0phdmFTY3JpcHRBdWRpb05vZGUuaWRsCkBAIC0yNywxMCArMjcs NyBAQCBtb2R1bGUgYXVkaW8gewogICAgIGludGVyZmFjZSBbCiAgICAgICAgIENvbmRpdGlvbmFs PVdFQl9BVURJTywKICAgICAgICAgR2VuZXJhdGVUb0pTLAotICAgICAgICBDdXN0b21NYXJrRnVu Y3Rpb24sCi0jaWYgZGVmaW5lZChWOF9CSU5ESU5HKSAmJiBWOF9CSU5ESU5HCi0gICAgICAgIEV2 ZW50VGFyZ2V0Ci0jZW5kaWYKKyAgICAgICAgQ3VzdG9tTWFya0Z1bmN0aW9uCiAgICAgXSBKYXZh U2NyaXB0QXVkaW9Ob2RlIDogQXVkaW9Ob2RlIHsKICAgICAgICAgLy8gUmVuZGVyaW5nIGNhbGxi YWNrCiAgICAgICAgIGF0dHJpYnV0ZSBFdmVudExpc3RlbmVyIG9uYXVkaW9wcm9jZXNzOwo= 106180 2011-09-02 13:04:57 -0700 2011-09-02 15:49:44 -0700 Patch bug-67516-20110902130456.patch text/plain 3175 dominicc U3VidmVyc2lvbiBSZXZpc2lvbjogOTQ0MzUKZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL0No YW5nZUxvZyBiL1NvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwppbmRleCA4MWZjMzQ5OWMwZGE3NjA4 ODIxMWNhMmVjMTc5ZWIwZWRkYTg4OWIyLi4yMzc0NTM1NDRlYWZhNmQyYmYwZDBmY2MzNWRkZDhm ZjEzYjkwZGFjIDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvU291 cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTYgQEAKKzIwMTEtMDktMDIgIERvbWlu aWMgQ29vbmV5ICA8ZG9taW5pY2NAY2hyb21pdW0ub3JnPgorCisgICAgICAgIE1ha2UgSmF2YVNj cmlwdEF1ZGlvTm9kZSBubyBsb25nZXIgY2xhaW0gdG8gYmUgYW4gRXZlbnRUYXJnZXQuIEl0IGlz IG5vdC4KKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTY3 NTE2CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAgVGVz dHM6IExheW91dFRlc3RzL3dlYmF1ZGlvLCBhbHRob3VnaCBKYXZhU2NyaXB0QXVkaW9Ob2RlIGlz IG5vdCB0ZXN0ZWQgOigKKworICAgICAgICAqIGJpbmRpbmdzL3NjcmlwdHMvQ29kZUdlbmVyYXRv clY4LnBtOgorICAgICAgICAoR2V0SW50ZXJuYWxGaWVsZHMpOiBKYXZhU2NyaXB0QXVkaW9Ob2Rl IHdhbnRzIHRvIGtlZXAgaXRzIG9uYXVkaW9wcm9jZXNzIGNhbGxiYWNrIGFsaXZlLgorICAgICAg ICAqIHdlYmF1ZGlvL0phdmFTY3JpcHRBdWRpb05vZGUuaWRsOiBEb24ndCBjbGFpbSB0byBiZSBF dmVudFRhcmdldC4KKwogMjAxMS0wOS0wMiAgU2FtIFdlaW5pZyAgPHNhbUB3ZWJraXQub3JnPgog CiAgICAgICAgIFJlbW92ZSBCZWZvcmVQcm9jZXNzRXZlbnQsIGl0IHdhcyBuZXZlciBtZWFudCB0 byBiZQpkaWZmIC0tZ2l0IGEvU291cmNlL1dlYkNvcmUvYmluZGluZ3MvanMvSlNFdmVudFRhcmdl dC5jcHAgYi9Tb3VyY2UvV2ViQ29yZS9iaW5kaW5ncy9qcy9KU0V2ZW50VGFyZ2V0LmNwcAppbmRl eCAyZjNmNjFiYTczMTdiM2U3NjU1NWVjY2Q4NmY1MjQwNzg3M2NhZDc4Li5jYTUzMDNkM2E5NmY2 MWM2ZDJhN2FjZGUyZDMwZTQwNTBiNmU2NmYxIDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViQ29yZS9i aW5kaW5ncy9qcy9KU0V2ZW50VGFyZ2V0LmNwcAorKysgYi9Tb3VyY2UvV2ViQ29yZS9iaW5kaW5n cy9qcy9KU0V2ZW50VGFyZ2V0LmNwcApAQCAtODYsOCArODYsNiBAQAogI2lmIEVOQUJMRShXRUJf QVVESU8pCiAjaW5jbHVkZSAiQXVkaW9Db250ZXh0LmgiCiAjaW5jbHVkZSAiSlNBdWRpb0NvbnRl eHQuaCIKLSNpbmNsdWRlICJKU0phdmFTY3JpcHRBdWRpb05vZGUuaCIKLSNpbmNsdWRlICJKYXZh U2NyaXB0QXVkaW9Ob2RlLmgiCiAjZW5kaWYKIAogI2lmIEVOQUJMRShXRUJfU09DS0VUUykKQEAg LTE3OCw4ICsxNzYsNiBAQCBKU1ZhbHVlIHRvSlMoRXhlY1N0YXRlKiBleGVjLCBKU0RPTUdsb2Jh bE9iamVjdCogZ2xvYmFsT2JqZWN0LCBFdmVudFRhcmdldCogdGFyZwogI2VuZGlmCiAKICNpZiBF TkFCTEUoV0VCX0FVRElPKQotICAgIGlmIChKYXZhU2NyaXB0QXVkaW9Ob2RlKiBqc0F1ZGlvTm9k ZSA9IHRhcmdldC0+dG9KYXZhU2NyaXB0QXVkaW9Ob2RlKCkpCi0gICAgICAgIHJldHVybiB0b0pT KGV4ZWMsIGdsb2JhbE9iamVjdCwganNBdWRpb05vZGUpOwogICAgIGlmIChBdWRpb0NvbnRleHQq IGF1ZGlvQ29udGV4dCA9IHRhcmdldC0+dG9BdWRpb0NvbnRleHQoKSkKICAgICAgICAgcmV0dXJu IHRvSlMoZXhlYywgZ2xvYmFsT2JqZWN0LCBhdWRpb0NvbnRleHQpOwogI2VuZGlmCmRpZmYgLS1n aXQgYS9Tb3VyY2UvV2ViQ29yZS9iaW5kaW5ncy9zY3JpcHRzL0NvZGVHZW5lcmF0b3JWOC5wbSBi L1NvdXJjZS9XZWJDb3JlL2JpbmRpbmdzL3NjcmlwdHMvQ29kZUdlbmVyYXRvclY4LnBtCmluZGV4 IDllMjYzNDA1Yjc0NTJkZWY0YmU1NGEzOTkwZjdlMWE5YTBmNmQ4ZTMuLjNiNWFjMWZhYmZiNTMy YzQ4NzVkYWEzOWMwZjU2MTBiODRiYmZmOGIgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XZWJDb3JlL2Jp bmRpbmdzL3NjcmlwdHMvQ29kZUdlbmVyYXRvclY4LnBtCisrKyBiL1NvdXJjZS9XZWJDb3JlL2Jp bmRpbmdzL3NjcmlwdHMvQ29kZUdlbmVyYXRvclY4LnBtCkBAIC01MDgsNiArNTA4LDcgQEAgc3Vi IEdldEludGVybmFsRmllbGRzCiAgICAgICAgICgkZGF0YU5vZGUtPmV4dGVuZGVkQXR0cmlidXRl cy0+eyJFdmVudFRhcmdldCJ9IHx8CiAgICAgICAgICBJc1N1YlR5cGUoJGRhdGFOb2RlLCAiQWJz dHJhY3RXb3JrZXIiKSB8fAogICAgICAgICAgSXNTdWJUeXBlKCRkYXRhTm9kZSwgIldvcmtlckNv bnRleHQiKSB8fAorICAgICAgICAgJG5hbWUgZXEgIkphdmFTY3JpcHRBdWRpb05vZGUiIHx8CiAg ICAgICAgICAkbmFtZSBlcSAiU1ZHRWxlbWVudEluc3RhbmNlIikpIHsKICAgICAgICAgcHVzaChA Y3VzdG9tSW50ZXJuYWxGaWVsZHMsICJldmVudExpc3RlbmVyQ2FjaGVJbmRleCIpOwogICAgIH0K ZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL3dlYmF1ZGlvL0phdmFTY3JpcHRBdWRpb05vZGUu aWRsIGIvU291cmNlL1dlYkNvcmUvd2ViYXVkaW8vSmF2YVNjcmlwdEF1ZGlvTm9kZS5pZGwKaW5k ZXggZWY1MzU5Yjk3NmVhYzYzMjFkMTg1YzkxZmE2OTUyMTVlZmU3ODk1OS4uYjEwMjg0OTlkNDQ4 MjJiMDlmYTBjZGNlMzYxMWQ5NmI4YTk1NDFkYSAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUv d2ViYXVkaW8vSmF2YVNjcmlwdEF1ZGlvTm9kZS5pZGwKKysrIGIvU291cmNlL1dlYkNvcmUvd2Vi YXVkaW8vSmF2YVNjcmlwdEF1ZGlvTm9kZS5pZGwKQEAgLTI3LDEwICsyNyw3IEBAIG1vZHVsZSBh dWRpbyB7CiAgICAgaW50ZXJmYWNlIFsKICAgICAgICAgQ29uZGl0aW9uYWw9V0VCX0FVRElPLAog ICAgICAgICBHZW5lcmF0ZVRvSlMsCi0gICAgICAgIEN1c3RvbU1hcmtGdW5jdGlvbiwKLSNpZiBk ZWZpbmVkKFY4X0JJTkRJTkcpICYmIFY4X0JJTkRJTkcKLSAgICAgICAgRXZlbnRUYXJnZXQKLSNl bmRpZgorICAgICAgICBDdXN0b21NYXJrRnVuY3Rpb24KICAgICBdIEphdmFTY3JpcHRBdWRpb05v ZGUgOiBBdWRpb05vZGUgewogICAgICAgICAvLyBSZW5kZXJpbmcgY2FsbGJhY2sKICAgICAgICAg YXR0cmlidXRlIEV2ZW50TGlzdGVuZXIgb25hdWRpb3Byb2Nlc3M7Cg==