151695 2015-11-30 22:34:56 -0800 Web Inspector: Split code coverage and type profilers 2026-01-12 09:04:56 -0800 1 1 1 Unclassified WebKit Web Inspector WebKit Local Build All All ASSIGNED InRadar P2 Normal --- 162469 160750 160973 146115 1 nvasilyev nvasilyev graouts inspector-bugzilla-changes saam webkit-bug-importer oldest_to_newest 1145142 0 nvasilyev 2015-11-30 22:34:56 -0800 * Split code coverage and type profilers; * Turn on code coverage profiler (grayed out text) by default; * Turn off type profiler (type info pills) by default; * [T] button will continue to toggle type profiler but won't affect code coverage profiler; * Code hotness visualizer, when implemented, will introduce a new button next to [T] and will be off by default. 1145143 1 webkit-bug-importer 2015-11-30 22:35:10 -0800 <rdar://problem/23700835> 1215437 2 284830 nvasilyev 2016-07-28 16:12:35 -0700 Created attachment 284830 WIP 1215438 3 284831 nvasilyev 2016-07-28 16:14:44 -0700 Created attachment 284831 WIP 1215441 4 bburg 2016-07-28 16:16:08 -0700 (In reply to comment #0) > * Split code coverage and type profilers; > * Turn on code coverage profiler (grayed out text) by default; > * Turn off type profiler (type info pills) by default; > * [T] button will continue to toggle type profiler but won't affect code > coverage profiler; > * Code hotness visualizer, when implemented, will introduce a new button > next to [T] and will be off by default. Please do not do all of these things in one patch. 1215444 5 joepeck 2016-07-28 16:19:06 -0700 (In reply to comment #0) > * Turn on code coverage profiler (grayed out text) by default; When doing this, it would be good to understand its performance impact. We spent a lot of time optimizing performance when the inspector is open for Timelines. It would be good to understand what the impact of this is if we have it always on. Or, change Timeline recording to also turn this off, to avoid a performance impact. 1215448 6 nvasilyev 2016-07-28 16:28:54 -0700 (In reply to comment #5) > (In reply to comment #0) > > * Turn on code coverage profiler (grayed out text) by default; > > When doing this, it would be good to understand its performance impact. We > spent a lot of time optimizing performance when the inspector is open for > Timelines. It would be good to understand what the impact of this is if we > have it always on. Or, change Timeline recording to also turn this off, to > avoid a performance impact. The front-end doesn't do anything performance intensive until a JS resource is visible. I don't know what happens on the backend. Saam should know. 1215451 7 nvasilyev 2016-07-28 16:31:20 -0700 (In reply to comment #4) > (In reply to comment #0) > > * Split code coverage and type profilers; > > * Turn on code coverage profiler (grayed out text) by default; > > * Turn off type profiler (type info pills) by default; > > * [T] button will continue to toggle type profiler but won't affect code > > coverage profiler; > > * Code hotness visualizer, when implemented, will introduce a new button > > next to [T] and will be off by default. > > Please do not do all of these things in one patch. I intended to do the first 4 points in one patch. 2nd to 4th points are just clarifications for the 1st. 1215455 8 joepeck 2016-07-28 16:33:31 -0700 (In reply to comment #6) > (In reply to comment #5) > > (In reply to comment #0) > > > * Turn on code coverage profiler (grayed out text) by default; > > > > When doing this, it would be good to understand its performance impact. We > > spent a lot of time optimizing performance when the inspector is open for > > Timelines. It would be good to understand what the impact of this is if we > > have it always on. Or, change Timeline recording to also turn this off, to > > avoid a performance impact. > > The front-end doesn't do anything performance intensive until a JS resource > is visible. > > I don't know what happens on the backend. Saam should know. I'm worried about how this will slow down content on the page. For example, running a benchmark with the code coverage enabled / not enabled. Anyone can run that and get numbers. It may be negligible, but we don't know until we measure. And we can't easily until this is separated from the Type Profiler, which you are doing! 1215460 9 nvasilyev 2016-07-28 16:47:02 -0700 Saam, RuntimeAgent.enableTypeProfiler() currently enables both type profiler AND basic blocks annotator (for grayed out unexecuted code). How hard would it be to decouple these two? Currently, calling RuntimeAgent.getBasicBlocks before RuntimeAgent.enableTypeProfiler() results in error: [Error] Error in getting basic block locations: The VM does not currently have a Control Flow Profiler. 1215966 10 saam 2016-07-31 14:24:34 -0700 (In reply to comment #0) > * Split code coverage and type profilers; This should be easy to do. It probably requires hacking on the protocol a bit, but the control flow profiler, and type profiler, are logically separate inside JSC. I can help review such a patch. I agree with Brian that this should be done in its own patch (I have less preference on the combining of other bits in other patches because I don't work on Inspector much). > * Turn on code coverage profiler (grayed out text) by default; This sounds like a good idea to me. The performance impact of control flow profiler was something like 10% last time I checked. Maybe a bit less. But regardless, as Joe mentioned, this should definitely be disabled when recording a timeline. We don't want this influencing the data we get out of the sampling profiler. > * Turn off type profiler (type info pills) by default; Isn't this already the default? > * [T] button will continue to toggle type profiler but won't affect code > coverage profiler; Sounds good to me. > * Code hotness visualizer, when implemented, will introduce a new button > next to [T] and will be off by default. What is the code hotness visualizer? Is this going to use data from the sampling profiler? Were you thinking this should be a hybrid of the sampling profiler and the control flow profiler? I think it's really important that it's not a hybrid. We don't want the control flow profiler to be enabled when sampling because it will bias the data we get from the sample. We worked really hard to have the inspector disable other JSC profilers/debugger features while taking a sample because we want JSC to run the program being sampled in as close to its natural state as possible. I guess if it's not a hybrid of these two profilers, it's important to ask if we really need to decouple this from the type profiler. It seems like there is both good and bad in doing this. It's nice to have the control flow profiler always on. It's bad to add complexity to the inspector's interface by allowing them to be decoupled. Others probably have stronger/more informed opinions about this than me. 1216177 11 nvasilyev 2016-08-01 11:34:20 -0700 (In reply to comment #10) > (In reply to comment #0) > > * Split code coverage and type profilers; > This should be easy to do. It probably requires hacking > on the protocol a bit, but the control flow profiler, and type profiler, > are logically separate inside JSC. I can help review such a patch. I agree > with Brian that this should be done in its own patch (I have less preference > on > the combining of other bits in other patches because I don't work on > Inspector much). I agree now that backend (JSC) and Web Inspector front-end changes should be in separate patches. > > > * Turn on code coverage profiler (grayed out text) by default; > This sounds like a good idea to me. The performance impact of control flow > profiler was something like 10% last time I checked. Maybe a bit less. > But regardless, as Joe mentioned, this should definitely be disabled when > recording a timeline. We don't want this influencing the data we get out > of the sampling profiler. Yes, it should be disabled when recording a timeline. > > > * Turn off type profiler (type info pills) by default; > Isn't this already the default? Yes, it is. I meant to write "Keep type profiler turned off by default". > > * Code hotness visualizer, when implemented, will introduce a new button > > next to [T] and will be off by default. > What is the code hotness visualizer? Is this going to use data from the > sampling > profiler? Were you thinking this should be a hybrid of the sampling profiler > and > the control flow profiler? I think it's really important that it's not a > hybrid. > We don't want the control flow profiler to be enabled when sampling because > it > will bias the data we get from the sample. We worked really hard to have the > inspector > disable other JSC profilers/debugger features while taking a sample because > we > want JSC to run the program being sampled in as close to its natural state > as possible. I had written this before sampling profiler was implemented. Let's keep the code hotness visualizer, whatever it would be, out of the scope of this bug. > > I guess if it's not a hybrid of these two profilers, it's important to ask > if we really > need to decouple this from the type profiler. It seems like there is both > good and bad > in doing this. It's nice to have the control flow profiler always on. It's > bad to add complexity > to the inspector's interface by allowing them to be decoupled. Others > probably have stronger/more > informed opinions about this than me. I find type profiler too intrusive to keep it enabled all the time. The control flow profiler adds a lot of value and doesn't take any extra screen space. 1218543 12 joepeck 2016-08-08 11:57:29 -0700 (In reply to comment #11) > (In reply to comment #10) > > (In reply to comment #0) > > > * Split code coverage and type profilers; > > [...] I agree with Brian that this should be done in its own patch > > (I have less preference on the combining of other bits in other > > patches because I don't work on Inspector much). > > I agree now that backend (JSC) and Web Inspector front-end changes should be > in separate patches. This is not so much about separate patches for backend and frontend changes. This is about separate patches for separate work. Splitting the two profilers is a single task, and that task should have its own bug and patch, separate from all the other tasks mentioned in this bug. 284830 2016-07-28 16:12:35 -0700 2016-07-28 16:14:44 -0700 WIP WIP.txt text/plain 4811 nvasilyev ZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJJbnNwZWN0b3JVSS9Vc2VySW50ZXJmYWNlL1ZpZXdzL1Nv dXJjZUNvZGVUZXh0RWRpdG9yLmpzIGIvU291cmNlL1dlYkluc3BlY3RvclVJL1VzZXJJbnRlcmZh Y2UvVmlld3MvU291cmNlQ29kZVRleHRFZGl0b3IuanMKaW5kZXggOWE5Njk3Mi4uMjk0NGI2MyAx MDA2NDQKLS0tIGEvU291cmNlL1dlYkluc3BlY3RvclVJL1VzZXJJbnRlcmZhY2UvVmlld3MvU291 cmNlQ29kZVRleHRFZGl0b3IuanMKKysrIGIvU291cmNlL1dlYkluc3BlY3RvclVJL1VzZXJJbnRl cmZhY2UvVmlld3MvU291cmNlQ29kZVRleHRFZGl0b3IuanMKQEAgLTEwOCw4ICsxMDgsMTEgQEAg V2ViSW5zcGVjdG9yLlNvdXJjZUNvZGVUZXh0RWRpdG9yID0gY2xhc3MgU291cmNlQ29kZVRleHRF ZGl0b3IgZXh0ZW5kcyBXZWJJbnNwZWMKICAgICAgICAgICAgIGlmICghdGhpcy5fdHlwZVRva2Vu U2Nyb2xsSGFuZGxlciAmJiAodGhpcy5fdHlwZVRva2VuQW5ub3RhdG9yIHx8IHRoaXMuX2Jhc2lj QmxvY2tBbm5vdGF0b3IpKQogICAgICAgICAgICAgICAgIHRoaXMuX2VuYWJsZVNjcm9sbEV2ZW50 c0ZvclR5cGVUb2tlbkFubm90YXRvcigpOwogICAgICAgICB9IGVsc2UgewotICAgICAgICAgICAg aWYgKHRoaXMuX3R5cGVUb2tlbkFubm90YXRvciB8fCB0aGlzLl9iYXNpY0Jsb2NrQW5ub3RhdG9y KQorICAgICAgICAgICAgaWYgKHRoaXMuX3R5cGVUb2tlbkFubm90YXRvcikKICAgICAgICAgICAg ICAgICB0aGlzLl9zZXRUeXBlVG9rZW5Bbm5vdGF0b3JFbmFibGVkU3RhdGUoZmFsc2UpOworCisg ICAgICAgICAgICBpZiAodGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvcikKKyAgICAgICAgICAgICAg ICB0aGlzLl9zZXRCYXNpY0Jsb2NrQW5ub3RhdG9yRW5hYmxlZFN0YXRlKGZhbHNlKQogICAgICAg ICB9CiAgICAgfQogCkBAIC0yNTEsMTAgKzI1NCwxNSBAQCBXZWJJbnNwZWN0b3IuU291cmNlQ29k ZVRleHRFZGl0b3IgPSBjbGFzcyBTb3VyY2VDb2RlVGV4dEVkaXRvciBleHRlbmRzIFdlYkluc3Bl YwogICAgICAgICBmb3IgKHZhciByYW5nZSBvZiBuZXdSYW5nZXMpCiAgICAgICAgICAgICB0aGlz Ll91cGRhdGVFZGl0YWJsZU1hcmtlcnMocmFuZ2UpOwogCi0gICAgICAgIGlmICh0aGlzLl90eXBl VG9rZW5Bbm5vdGF0b3IgfHwgdGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvcikgeworICAgICAgICBp ZiAodGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvcikgeworICAgICAgICAgICAgdGhpcy5fc2V0QmFz aWNCbG9ja0Fubm90YXRvckVuYWJsZWRTdGF0ZShmYWxzZSk7CisgICAgICAgICAgICB0aGlzLl9i YXNpY0Jsb2NrQW5ub3RhdG9yID0gbnVsbDsKKyAgICAgICAgfQorCisgICAgICAgIGlmICh0aGlz Ll90eXBlVG9rZW5Bbm5vdGF0b3IpIHsKICAgICAgICAgICAgIHRoaXMuX3NldFR5cGVUb2tlbkFu bm90YXRvckVuYWJsZWRTdGF0ZShmYWxzZSk7CiAgICAgICAgICAgICB0aGlzLl90eXBlVG9rZW5B bm5vdGF0b3IgPSBudWxsOwotICAgICAgICAgICAgdGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvciA9 IG51bGw7CisKICAgICAgICAgfQogICAgIH0KIApAQCAtMzAxLDE4ICszMDksMjggQEAgV2ViSW5z cGVjdG9yLlNvdXJjZUNvZGVUZXh0RWRpdG9yID0gY2xhc3MgU291cmNlQ29kZVRleHRFZGl0b3Ig ZXh0ZW5kcyBXZWJJbnNwZWMKICAgICAgICAgLy8gVGhlIGFubm90YXRvcnMgbXVzdCBiZSBjbGVh cmVkIGJlZm9yZSBwcmV0dHkgcHJpbnRpbmcgdGFrZXMgcGxhY2UgYW5kIHJlc3VtZWQKICAgICAg ICAgLy8gYWZ0ZXIgc28gdGhhdCB0aGV5IGNsZWFyIHRoZWlyIGFubm90YXRpb25zIGluIGEga25v d24gc3RhdGUgYW5kIGluc2VydCBuZXcgYW5ub3RhdGlvbnMKICAgICAgICAgLy8gaW4gdGhlIG5l dyBzdGF0ZS4KLSAgICAgICAgdmFyIHNob3VsZFJlc3VtZVR5cGVUb2tlbkFubm90YXRvciA9IHRo aXMuX3R5cGVUb2tlbkFubm90YXRvciAmJiB0aGlzLl90eXBlVG9rZW5Bbm5vdGF0b3IuaXNBY3Rp dmUoKTsKKwogICAgICAgICB2YXIgc2hvdWxkUmVzdW1lQmFzaWNCbG9ja0Fubm90YXRvciA9IHRo aXMuX2Jhc2ljQmxvY2tBbm5vdGF0b3IgJiYgdGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvci5pc0Fj dGl2ZSgpOwotICAgICAgICBpZiAoc2hvdWxkUmVzdW1lVHlwZVRva2VuQW5ub3RhdG9yIHx8IHNo b3VsZFJlc3VtZUJhc2ljQmxvY2tBbm5vdGF0b3IpCisgICAgICAgIGlmIChzaG91bGRSZXN1bWVC YXNpY0Jsb2NrQW5ub3RhdG9yKQorICAgICAgICAgICAgdGhpcy5fc2V0QmFzaWNCbG9ja0Fubm90 YXRvckVuYWJsZWRTdGF0ZShmYWxzZSk7CisKKyAgICAgICAgdmFyIHNob3VsZFJlc3VtZVR5cGVU b2tlbkFubm90YXRvciA9IHRoaXMuX3R5cGVUb2tlbkFubm90YXRvciAmJiB0aGlzLl90eXBlVG9r ZW5Bbm5vdGF0b3IuaXNBY3RpdmUoKTsKKyAgICAgICAgaWYgKHNob3VsZFJlc3VtZVR5cGVUb2tl bkFubm90YXRvcikKICAgICAgICAgICAgIHRoaXMuX3NldFR5cGVUb2tlbkFubm90YXRvckVuYWJs ZWRTdGF0ZShmYWxzZSk7CiAKICAgICAgICAgcmV0dXJuIHN1cGVyLnByZXR0eVByaW50KHByZXR0 eSkudGhlbigoKSA9PiB7CiAgICAgICAgICAgICBpZiAocHJldHR5IHx8ICF0aGlzLl9pc1Byb2Jh Ymx5TWluaWZpZWQpIHsKLSAgICAgICAgICAgICAgICBpZiAoc2hvdWxkUmVzdW1lVHlwZVRva2Vu QW5ub3RhdG9yIHx8IHNob3VsZFJlc3VtZUJhc2ljQmxvY2tBbm5vdGF0b3IpCisgICAgICAgICAg ICAgICAgaWYgKHNob3VsZFJlc3VtZUJhc2ljQmxvY2tBbm5vdGF0b3IpCisgICAgICAgICAgICAg ICAgICAgIHRoaXMuX3NldEJhc2ljQmxvY2tBbm5vdGF0b3JFbmFibGVkU3RhdGUodHJ1ZSk7CisK KyAgICAgICAgICAgICAgICBpZiAoc2hvdWxkUmVzdW1lVHlwZVRva2VuQW5ub3RhdG9yKQogICAg ICAgICAgICAgICAgICAgICB0aGlzLl9zZXRUeXBlVG9rZW5Bbm5vdGF0b3JFbmFibGVkU3RhdGUo dHJ1ZSk7CiAgICAgICAgICAgICB9IGVsc2UgewogICAgICAgICAgICAgICAgIGNvbnNvbGUuYXNz ZXJ0KCFwcmV0dHkgJiYgdGhpcy5faXNQcm9iYWJseU1pbmlmaWVkKTsKLSAgICAgICAgICAgICAg ICBpZiAodGhpcy5fdHlwZVRva2VuQW5ub3RhdG9yIHx8IHRoaXMuX2Jhc2ljQmxvY2tBbm5vdGF0 b3IpCisgICAgICAgICAgICAgICAgaWYgKHRoaXMuX2Jhc2ljQmxvY2tBbm5vdGF0b3IpCisgICAg ICAgICAgICAgICAgICAgIHRoaXMuX3NldEJhc2ljQmxvY2tBbm5vdGF0b3JFbmFibGVkU3RhdGUo ZmFsc2UpOworCisgICAgICAgICAgICAgICAgaWYgKHRoaXMuX3R5cGVUb2tlbkFubm90YXRvcikK ICAgICAgICAgICAgICAgICAgICAgdGhpcy5fc2V0VHlwZVRva2VuQW5ub3RhdG9yRW5hYmxlZFN0 YXRlKGZhbHNlKTsKICAgICAgICAgICAgIH0KICAgICAgICAgfSk7CkBAIC00MTQsMTEgKzQzMiwx NCBAQCBXZWJJbnNwZWN0b3IuU291cmNlQ29kZVRleHRFZGl0b3IgPSBjbGFzcyBTb3VyY2VDb2Rl VGV4dEVkaXRvciBleHRlbmRzIFdlYkluc3BlYwogCiAgICAgICAgIHRoaXMuc3RyaW5nID0gY29u dGVudDsKIAotICAgICAgICB0aGlzLl9tYWtlVHlwZVRva2VuQW5ub3RhdG9yKCk7CiAgICAgICAg IHRoaXMuX21ha2VCYXNpY0Jsb2NrQW5ub3RhdG9yKCk7CisgICAgICAgIGlmICh0aGlzLl9iYXNp Y0Jsb2NrQW5ub3RhdG9yKQorICAgICAgICAgICAgdGhpcy5fc2V0QmFzaWNCbG9ja0Fubm90YXRv ckVuYWJsZWRTdGF0ZSh0cnVlKTsKIAorICAgICAgICAvL0ZJWEVECiAgICAgICAgIGlmIChXZWJJ bnNwZWN0b3Iuc2hvd0phdmFTY3JpcHRUeXBlSW5mb3JtYXRpb25TZXR0aW5nLnZhbHVlKSB7Ci0g ICAgICAgICAgICBpZiAodGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvciB8fCB0aGlzLl90eXBlVG9r ZW5Bbm5vdGF0b3IpCisgICAgICAgICAgICB0aGlzLl9tYWtlVHlwZVRva2VuQW5ub3RhdG9yKCk7 CisgICAgICAgICAgICBpZiAodGhpcy5fdHlwZVRva2VuQW5ub3RhdG9yKQogICAgICAgICAgICAg ICAgIHRoaXMuX3NldFR5cGVUb2tlbkFubm90YXRvckVuYWJsZWRTdGF0ZSh0cnVlKTsKICAgICAg ICAgfQogCkBAIC0xNzIzLDYgKzE3NDQsMjEgQEAgV2ViSW5zcGVjdG9yLlNvdXJjZUNvZGVUZXh0 RWRpdG9yID0gY2xhc3MgU291cmNlQ29kZVRleHRFZGl0b3IgZXh0ZW5kcyBXZWJJbnNwZWMKICAg ICAgICAgdGhpcy5fdXBkYXRlVG9rZW5UcmFja2luZ0NvbnRyb2xsZXJTdGF0ZSgpOwogICAgIH0K IAorICAgIF9zZXRCYXNpY0Jsb2NrQW5ub3RhdG9yRW5hYmxlZFN0YXRlKHNob3VsZEFjdGl2YXRl KQorICAgIHsKKyAgICAgICAgaWYgKHNob3VsZEFjdGl2YXRlKSB7CisgICAgICAgICAgICBjb25z b2xlLmFzc2VydCh0aGlzLnZpc2libGUsICJBbm5vdGF0b3JzIHNob3VsZCBub3QgYmUgZW5hYmxl ZCBpZiB0aGUgVGV4dEVkaXRvciBpcyBub3QgdmlzaWJsZSIpOworCisgICAgICAgICAgICBpZiAo dGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvcikgeworICAgICAgICAgICAgICAgIGNvbnNvbGUuYXNz ZXJ0KCF0aGlzLl9iYXNpY0Jsb2NrQW5ub3RhdG9yLmlzQWN0aXZlKCkpOworICAgICAgICAgICAg ICAgIHRoaXMuX2Jhc2ljQmxvY2tBbm5vdGF0b3IucmVzZXQoKTsKKyAgICAgICAgICAgIH0KKyAg ICAgICAgfSBlbHNlIHsKKyAgICAgICAgICAgIGlmICh0aGlzLl9iYXNpY0Jsb2NrQW5ub3RhdG9y KQorICAgICAgICAgICAgICAgIHRoaXMuX2Jhc2ljQmxvY2tBbm5vdGF0b3IuY2xlYXIoKTsKKyAg ICAgICAgfQorICAgIH0KKwogICAgIF9nZXRBc3NvY2lhdGVkU2NyaXB0KCkKICAgICB7CiAgICAg ICAgIHZhciBzY3JpcHQgPSBudWxsOwo= 284831 2016-07-28 16:14:44 -0700 2016-07-28 16:14:44 -0700 WIP WIP.txt text/plain 4792 nvasilyev ZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJJbnNwZWN0b3JVSS9Vc2VySW50ZXJmYWNlL1ZpZXdzL1Nv dXJjZUNvZGVUZXh0RWRpdG9yLmpzIGIvU291cmNlL1dlYkluc3BlY3RvclVJL1VzZXJJbnRlcmZh Y2UvVmlld3MvU291cmNlQ29kZVRleHRFZGl0b3IuanMKaW5kZXggOWE5Njk3Mi4uYzFlMTUzYyAx MDA2NDQKLS0tIGEvU291cmNlL1dlYkluc3BlY3RvclVJL1VzZXJJbnRlcmZhY2UvVmlld3MvU291 cmNlQ29kZVRleHRFZGl0b3IuanMKKysrIGIvU291cmNlL1dlYkluc3BlY3RvclVJL1VzZXJJbnRl cmZhY2UvVmlld3MvU291cmNlQ29kZVRleHRFZGl0b3IuanMKQEAgLTEwOCw4ICsxMDgsMTEgQEAg V2ViSW5zcGVjdG9yLlNvdXJjZUNvZGVUZXh0RWRpdG9yID0gY2xhc3MgU291cmNlQ29kZVRleHRF ZGl0b3IgZXh0ZW5kcyBXZWJJbnNwZWMKICAgICAgICAgICAgIGlmICghdGhpcy5fdHlwZVRva2Vu U2Nyb2xsSGFuZGxlciAmJiAodGhpcy5fdHlwZVRva2VuQW5ub3RhdG9yIHx8IHRoaXMuX2Jhc2lj QmxvY2tBbm5vdGF0b3IpKQogICAgICAgICAgICAgICAgIHRoaXMuX2VuYWJsZVNjcm9sbEV2ZW50 c0ZvclR5cGVUb2tlbkFubm90YXRvcigpOwogICAgICAgICB9IGVsc2UgewotICAgICAgICAgICAg aWYgKHRoaXMuX3R5cGVUb2tlbkFubm90YXRvciB8fCB0aGlzLl9iYXNpY0Jsb2NrQW5ub3RhdG9y KQorICAgICAgICAgICAgaWYgKHRoaXMuX3R5cGVUb2tlbkFubm90YXRvcikKICAgICAgICAgICAg ICAgICB0aGlzLl9zZXRUeXBlVG9rZW5Bbm5vdGF0b3JFbmFibGVkU3RhdGUoZmFsc2UpOworCisg ICAgICAgICAgICBpZiAodGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvcikKKyAgICAgICAgICAgICAg ICB0aGlzLl9zZXRCYXNpY0Jsb2NrQW5ub3RhdG9yRW5hYmxlZFN0YXRlKGZhbHNlKQogICAgICAg ICB9CiAgICAgfQogCkBAIC0yNTEsMTAgKzI1NCwxNCBAQCBXZWJJbnNwZWN0b3IuU291cmNlQ29k ZVRleHRFZGl0b3IgPSBjbGFzcyBTb3VyY2VDb2RlVGV4dEVkaXRvciBleHRlbmRzIFdlYkluc3Bl YwogICAgICAgICBmb3IgKHZhciByYW5nZSBvZiBuZXdSYW5nZXMpCiAgICAgICAgICAgICB0aGlz Ll91cGRhdGVFZGl0YWJsZU1hcmtlcnMocmFuZ2UpOwogCi0gICAgICAgIGlmICh0aGlzLl90eXBl VG9rZW5Bbm5vdGF0b3IgfHwgdGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvcikgeworICAgICAgICBp ZiAodGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvcikgeworICAgICAgICAgICAgdGhpcy5fc2V0QmFz aWNCbG9ja0Fubm90YXRvckVuYWJsZWRTdGF0ZShmYWxzZSk7CisgICAgICAgICAgICB0aGlzLl9i YXNpY0Jsb2NrQW5ub3RhdG9yID0gbnVsbDsKKyAgICAgICAgfQorCisgICAgICAgIGlmICh0aGlz Ll90eXBlVG9rZW5Bbm5vdGF0b3IpIHsKICAgICAgICAgICAgIHRoaXMuX3NldFR5cGVUb2tlbkFu bm90YXRvckVuYWJsZWRTdGF0ZShmYWxzZSk7CiAgICAgICAgICAgICB0aGlzLl90eXBlVG9rZW5B bm5vdGF0b3IgPSBudWxsOwotICAgICAgICAgICAgdGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvciA9 IG51bGw7CiAgICAgICAgIH0KICAgICB9CiAKQEAgLTMwMSwxOCArMzA4LDI4IEBAIFdlYkluc3Bl Y3Rvci5Tb3VyY2VDb2RlVGV4dEVkaXRvciA9IGNsYXNzIFNvdXJjZUNvZGVUZXh0RWRpdG9yIGV4 dGVuZHMgV2ViSW5zcGVjCiAgICAgICAgIC8vIFRoZSBhbm5vdGF0b3JzIG11c3QgYmUgY2xlYXJl ZCBiZWZvcmUgcHJldHR5IHByaW50aW5nIHRha2VzIHBsYWNlIGFuZCByZXN1bWVkCiAgICAgICAg IC8vIGFmdGVyIHNvIHRoYXQgdGhleSBjbGVhciB0aGVpciBhbm5vdGF0aW9ucyBpbiBhIGtub3du IHN0YXRlIGFuZCBpbnNlcnQgbmV3IGFubm90YXRpb25zCiAgICAgICAgIC8vIGluIHRoZSBuZXcg c3RhdGUuCi0gICAgICAgIHZhciBzaG91bGRSZXN1bWVUeXBlVG9rZW5Bbm5vdGF0b3IgPSB0aGlz Ll90eXBlVG9rZW5Bbm5vdGF0b3IgJiYgdGhpcy5fdHlwZVRva2VuQW5ub3RhdG9yLmlzQWN0aXZl KCk7CisKICAgICAgICAgdmFyIHNob3VsZFJlc3VtZUJhc2ljQmxvY2tBbm5vdGF0b3IgPSB0aGlz Ll9iYXNpY0Jsb2NrQW5ub3RhdG9yICYmIHRoaXMuX2Jhc2ljQmxvY2tBbm5vdGF0b3IuaXNBY3Rp dmUoKTsKLSAgICAgICAgaWYgKHNob3VsZFJlc3VtZVR5cGVUb2tlbkFubm90YXRvciB8fCBzaG91 bGRSZXN1bWVCYXNpY0Jsb2NrQW5ub3RhdG9yKQorICAgICAgICBpZiAoc2hvdWxkUmVzdW1lQmFz aWNCbG9ja0Fubm90YXRvcikKKyAgICAgICAgICAgIHRoaXMuX3NldEJhc2ljQmxvY2tBbm5vdGF0 b3JFbmFibGVkU3RhdGUoZmFsc2UpOworCisgICAgICAgIHZhciBzaG91bGRSZXN1bWVUeXBlVG9r ZW5Bbm5vdGF0b3IgPSB0aGlzLl90eXBlVG9rZW5Bbm5vdGF0b3IgJiYgdGhpcy5fdHlwZVRva2Vu QW5ub3RhdG9yLmlzQWN0aXZlKCk7CisgICAgICAgIGlmIChzaG91bGRSZXN1bWVUeXBlVG9rZW5B bm5vdGF0b3IpCiAgICAgICAgICAgICB0aGlzLl9zZXRUeXBlVG9rZW5Bbm5vdGF0b3JFbmFibGVk U3RhdGUoZmFsc2UpOwogCiAgICAgICAgIHJldHVybiBzdXBlci5wcmV0dHlQcmludChwcmV0dHkp LnRoZW4oKCkgPT4gewogICAgICAgICAgICAgaWYgKHByZXR0eSB8fCAhdGhpcy5faXNQcm9iYWJs eU1pbmlmaWVkKSB7Ci0gICAgICAgICAgICAgICAgaWYgKHNob3VsZFJlc3VtZVR5cGVUb2tlbkFu bm90YXRvciB8fCBzaG91bGRSZXN1bWVCYXNpY0Jsb2NrQW5ub3RhdG9yKQorICAgICAgICAgICAg ICAgIGlmIChzaG91bGRSZXN1bWVCYXNpY0Jsb2NrQW5ub3RhdG9yKQorICAgICAgICAgICAgICAg ICAgICB0aGlzLl9zZXRCYXNpY0Jsb2NrQW5ub3RhdG9yRW5hYmxlZFN0YXRlKHRydWUpOworCisg ICAgICAgICAgICAgICAgaWYgKHNob3VsZFJlc3VtZVR5cGVUb2tlbkFubm90YXRvcikKICAgICAg ICAgICAgICAgICAgICAgdGhpcy5fc2V0VHlwZVRva2VuQW5ub3RhdG9yRW5hYmxlZFN0YXRlKHRy dWUpOwogICAgICAgICAgICAgfSBlbHNlIHsKICAgICAgICAgICAgICAgICBjb25zb2xlLmFzc2Vy dCghcHJldHR5ICYmIHRoaXMuX2lzUHJvYmFibHlNaW5pZmllZCk7Ci0gICAgICAgICAgICAgICAg aWYgKHRoaXMuX3R5cGVUb2tlbkFubm90YXRvciB8fCB0aGlzLl9iYXNpY0Jsb2NrQW5ub3RhdG9y KQorICAgICAgICAgICAgICAgIGlmICh0aGlzLl9iYXNpY0Jsb2NrQW5ub3RhdG9yKQorICAgICAg ICAgICAgICAgICAgICB0aGlzLl9zZXRCYXNpY0Jsb2NrQW5ub3RhdG9yRW5hYmxlZFN0YXRlKGZh bHNlKTsKKworICAgICAgICAgICAgICAgIGlmICh0aGlzLl90eXBlVG9rZW5Bbm5vdGF0b3IpCiAg ICAgICAgICAgICAgICAgICAgIHRoaXMuX3NldFR5cGVUb2tlbkFubm90YXRvckVuYWJsZWRTdGF0 ZShmYWxzZSk7CiAgICAgICAgICAgICB9CiAgICAgICAgIH0pOwpAQCAtNDE0LDExICs0MzEsMTMg QEAgV2ViSW5zcGVjdG9yLlNvdXJjZUNvZGVUZXh0RWRpdG9yID0gY2xhc3MgU291cmNlQ29kZVRl eHRFZGl0b3IgZXh0ZW5kcyBXZWJJbnNwZWMKIAogICAgICAgICB0aGlzLnN0cmluZyA9IGNvbnRl bnQ7CiAKLSAgICAgICAgdGhpcy5fbWFrZVR5cGVUb2tlbkFubm90YXRvcigpOwogICAgICAgICB0 aGlzLl9tYWtlQmFzaWNCbG9ja0Fubm90YXRvcigpOworICAgICAgICBpZiAodGhpcy5fYmFzaWNC bG9ja0Fubm90YXRvcikKKyAgICAgICAgICAgIHRoaXMuX3NldEJhc2ljQmxvY2tBbm5vdGF0b3JF bmFibGVkU3RhdGUodHJ1ZSk7CiAKICAgICAgICAgaWYgKFdlYkluc3BlY3Rvci5zaG93SmF2YVNj cmlwdFR5cGVJbmZvcm1hdGlvblNldHRpbmcudmFsdWUpIHsKLSAgICAgICAgICAgIGlmICh0aGlz Ll9iYXNpY0Jsb2NrQW5ub3RhdG9yIHx8IHRoaXMuX3R5cGVUb2tlbkFubm90YXRvcikKKyAgICAg ICAgICAgIHRoaXMuX21ha2VUeXBlVG9rZW5Bbm5vdGF0b3IoKTsKKyAgICAgICAgICAgIGlmICh0 aGlzLl90eXBlVG9rZW5Bbm5vdGF0b3IpCiAgICAgICAgICAgICAgICAgdGhpcy5fc2V0VHlwZVRv a2VuQW5ub3RhdG9yRW5hYmxlZFN0YXRlKHRydWUpOwogICAgICAgICB9CiAKQEAgLTE3MjMsNiAr MTc0MiwyMSBAQCBXZWJJbnNwZWN0b3IuU291cmNlQ29kZVRleHRFZGl0b3IgPSBjbGFzcyBTb3Vy Y2VDb2RlVGV4dEVkaXRvciBleHRlbmRzIFdlYkluc3BlYwogICAgICAgICB0aGlzLl91cGRhdGVU b2tlblRyYWNraW5nQ29udHJvbGxlclN0YXRlKCk7CiAgICAgfQogCisgICAgX3NldEJhc2ljQmxv Y2tBbm5vdGF0b3JFbmFibGVkU3RhdGUoc2hvdWxkQWN0aXZhdGUpCisgICAgeworICAgICAgICBp ZiAoc2hvdWxkQWN0aXZhdGUpIHsKKyAgICAgICAgICAgIGNvbnNvbGUuYXNzZXJ0KHRoaXMudmlz aWJsZSwgIkFubm90YXRvcnMgc2hvdWxkIG5vdCBiZSBlbmFibGVkIGlmIHRoZSBUZXh0RWRpdG9y IGlzIG5vdCB2aXNpYmxlIik7CisKKyAgICAgICAgICAgIGlmICh0aGlzLl9iYXNpY0Jsb2NrQW5u b3RhdG9yKSB7CisgICAgICAgICAgICAgICAgY29uc29sZS5hc3NlcnQoIXRoaXMuX2Jhc2ljQmxv Y2tBbm5vdGF0b3IuaXNBY3RpdmUoKSk7CisgICAgICAgICAgICAgICAgdGhpcy5fYmFzaWNCbG9j a0Fubm90YXRvci5yZXNldCgpOworICAgICAgICAgICAgfQorICAgICAgICB9IGVsc2UgeworICAg ICAgICAgICAgaWYgKHRoaXMuX2Jhc2ljQmxvY2tBbm5vdGF0b3IpCisgICAgICAgICAgICAgICAg dGhpcy5fYmFzaWNCbG9ja0Fubm90YXRvci5jbGVhcigpOworICAgICAgICB9CisgICAgfQorCiAg ICAgX2dldEFzc29jaWF0ZWRTY3JpcHQoKQogICAgIHsKICAgICAgICAgdmFyIHNjcmlwdCA9IG51 bGw7Cg==