100873 2012-10-31 11:41:02 -0700 REGRESSION (r132941): attribute modification 10% performance regression 2012-11-01 18:12:01 -0700 1 1 1 Unclassified WebKit DOM 528+ (Nightly build) Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 ojan webkit-unassigned abarth allan.jensen cmarcelo dglazkov kling koivisto macpherson menard webkit.review.bot oldest_to_newest 755499 0 ojan 2012-10-31 11:41:02 -0700 Regression range: http://trac.webkit.org/log/?verbose=on&rev=132949&stop_rev=132924 http://trac.webkit.org/changeset/132941/ seems like the most likely culprit to me. Most clear graph. Click on the part in the graph where the performance drops then click on the webkit link on the right to see the webkit blamelist: http://build.chromium.org/f/chromium/perf/chromium-rel-win7-webkit/dromaeo_jslibattrjquery/report.html?rev=165231&graph=jslib_attr_jquery_jQuery___removeClass&trace=score&history=150 Some less clear graphs that point at the same webkit blamelist: http://build.chromium.org/f/chromium/perf/chromium-rel-win7-webkit/dromaeo_jslibattrprototype/report.html?rev=165213&graph=jslib_attr_prototype_Prototype___removeClassName&trace=score&history=150 http://build.chromium.org/f/chromium/perf/chromium-rel-win7-webkit/dromaeo_jslibattrprototype/report.html?rev=165213&graph=jslib_attr_prototype_Prototype___writeAttribute&trace=score&history=150 http://build.chromium.org/f/chromium/perf/chromium-rel-win7-webkit/dromaeo_jslibattrjquery/report.html?rev=165213&graph=jslib_attr_jquery_jQuery___attr_class_test_&trace=score&history=150 755506 1 ojan 2012-10-31 11:48:45 -0700 Smaller regression range from the Chromium linux bot: http://trac.webkit.org/log/?verbose=on&rev=132944&stop_rev=132938 http://build.chromium.org/f/chromium/perf/linux-release-webkit-latest/dromaeo_jslibattrjquery/report.html?rev=165139&graph=jslib_attr_jquery_jQuery___attr_class_test_&trace=score&history=150 http://build.chromium.org/f/chromium/perf/linux-release-webkit-latest/dromaeo_jslibattrprototype/report.html?rev=165139&graph=jslib_attr_prototype_Prototype___removeClassName&trace=score&history=150 755524 2 kling 2012-10-31 12:04:06 -0700 If it was http://trac.webkit.org/changeset/132941, I think http://trac.webkit.org/changeset/133021 should fix it. 755561 3 ojan 2012-10-31 12:48:18 -0700 (In reply to comment #2) > If it was http://trac.webkit.org/changeset/132941, I think http://trac.webkit.org/changeset/133021 should fix it. 133021 did not fix it. 755562 4 koivisto 2012-10-31 12:51:39 -0700 Not sure it will. We'll see. I think a sufficiently artificial test case might see a regression from that change. There are ways to improve if needed. 755575 5 koivisto 2012-10-31 13:10:35 -0700 Turns out the difference finding algorithm using SpaceSplitString mutation is too slow: Running Time Self Symbol Name 1714.0ms 5.6% 302,0 WebCore::Element::classAttributeChanged(WTF::AtomicString const&) 532.0ms 1.7% 174,0 WebCore::SpaceSplitString::remove(WTF::AtomicString const&) 277.0ms 0.9% 66,0 WebCore::SpaceSplitString::ensureUnique() 136.0ms 0.4% 136,0 WTF::fastMalloc(unsigned long) 75.0ms 0.2% 75,0 WTF::Vector<WTF::AtomicString, 4ul>::Vector(WTF::Vector<WTF::AtomicString, 4ul> const&) Should be easy to optimize this away. 756447 6 171893 koivisto 2012-11-01 10:50:51 -0700 Created attachment 171893 patch 756481 7 171893 ojan 2012-11-01 11:31:40 -0700 Comment on attachment 171893 patch View in context: https://bugs.webkit.org/attachment.cgi?id=171893&action=review > Source/WebCore/dom/Element.cpp:723 > + StyleResolver* styleResolver = document()->styleResolverIfExists(); > + bool testShouldInvalidateStyle = attached() && styleResolver && styleChangeType() < FullStyleChange; I almost wonder if it's worth adding a static helper function here since this code is duplicated with Element::classAttributeChanged. static StyleResolver* styleResolverForTestingStyleInvalidation(Document* document, Element* element) { if (!element->attached() || element->needsStyleRecalc() || styleTypeChange >= FullStyleChange) return 0; return document()->styleResolverIfExists(); } Then you wouldn't need the testShouldInvalidateStyle at all and could just null check the returned style resolver. It's not clearly better to me, but it does avoid some duplication. Up to you. > Source/WebCore/dom/Element.cpp:733 > + if (shouldInvalidateStyle) > + testShouldInvalidateStyle = false; I think you could avoid needing to do this if you put the check below into an else clause. That'd simplify the logic some I think. } else { if (name == HTMLNames::nameAttr) setHasName(!newValue.isNull()); shouldInvalidateStyle = ...; } > Source/WebCore/dom/Element.cpp:738 > + shouldInvalidateStyle = testShouldInvalidateStyle && styleResolver->hasSelectorForAttribute(name.localName()); Not sure it matters, but we currently only do this if (!needsStyleRecalc()). Maybe testShouldInvalidateStyle should also && !needsStyleRecalc()? Obviously the same would apply in Element::classAttributeChanged. Saves a hashtable lookup. Not sure it matters in practice. 756501 8 koivisto 2012-11-01 11:45:05 -0700 (In reply to comment #7) > Not sure it matters, but we currently only do this if (!needsStyleRecalc()). Maybe testShouldInvalidateStyle should also && !needsStyleRecalc()? Obviously the same would apply in Element::classAttributeChanged. Saves a hashtable lookup. Not sure it matters in practice. styleChangeType() < FullStyleChange is the equivalent test (more correct in fact). 756511 9 ojan 2012-11-01 11:54:38 -0700 (In reply to comment #8) > styleChangeType() < FullStyleChange is the equivalent test (more correct in fact). Ah. Got it. 756527 10 171893 webkit.review.bot 2012-11-01 12:07:39 -0700 Comment on attachment 171893 patch Attachment 171893 did not pass chromium-ews (chromium-xvfb): Output: http://queues.webkit.org/results/14686598 New failing tests: css3/viewport-percentage-lengths/css3-viewport-percentage-lengths-getStyle.html fast/dynamic/004.html 756541 11 koivisto 2012-11-01 12:41:53 -0700 http://trac.webkit.org/changeset/133214 (with the test breaking bug fixed) 756543 12 koivisto 2012-11-01 12:42:13 -0700 Lets see what the bots say. 756719 13 koivisto 2012-11-01 16:31:21 -0700 Looks like it fixed it. http://build.chromium.org/f/chromium/perf/chromium-rel-win7-webkit/dromaeo_jslibattrjquery/report.html?graph=jslib_attr_jquery_jQuery___removeClass&trace=score&history=100&rev=165526 756778 14 ojan 2012-11-01 18:12:01 -0700 Yup. Looks great! 171893 2012-11-01 10:50:51 -0700 2012-11-01 12:07:39 -0700 patch classAttributeChanged-performance.patch text/plain 10754 koivisto SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDEzMzE5MCkKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDM4IEBACisyMDEyLTExLTAxICBBbnR0aSBL b2l2aXN0byAgPGFudHRpQGFwcGxlLmNvbT4KKworICAgICAgICBSRUdSRVNTSU9OIChyMTMyOTQx KTogYXR0cmlidXRlIG1vZGlmaWNhdGlvbiAxMCUgcGVyZm9ybWFuY2UgcmVncmVzc2lvbgorICAg ICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MTAwODczCisKKyAg ICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAgRG9uJ3QgZG8gdGhl IGNsYXNzIGNoYW5nZSBmaW5kaW5nIGJ5IG11dGF0aW5nIFNwYWNlU3BsaXRTdHJpbmcuIEl0IGlz IHNsb3cuIEluc3RlYWQgdXNlIGEgYml0IHZlY3RvcgorICAgICAgICB0byBtYXJrIHRoZSB1bmNo YW5nZWQgY2xhc3NlcworCisgICAgICAgICogY3NzL1N0eWxlUmVzb2x2ZXIuY3BwOgorICAgICAg ICAoV2ViQ29yZSk6CisgICAgICAgICogY3NzL1N0eWxlUmVzb2x2ZXIuaDoKKyAgICAgICAgKFdl YkNvcmU6OlN0eWxlUmVzb2x2ZXI6Omhhc1NlbGVjdG9yRm9yQXR0cmlidXRlKToKKyAgICAgICAg KFdlYkNvcmUpOgorICAgICAgICAoV2ViQ29yZTo6U3R5bGVSZXNvbHZlcjo6aGFzU2VsZWN0b3JG b3JDbGFzcyk6CisgICAgICAgIChXZWJDb3JlOjpTdHlsZVJlc29sdmVyOjpoYXNTZWxlY3RvckZv cklkKToKKyAgICAgICAgCisgICAgICAgICAgICBJbmxpbmVkIHRoZXNlIGFuZCBtb3ZlZCB2YWx1 ZSB2YWxpZGl0eSB0ZXN0aW5nIHRvIGNsaWVudHMuCisKKyAgICAgICAgKiBkb20vRWxlbWVudC5j cHA6CisgICAgICAgIChXZWJDb3JlOjpjaGVja05lZWRzU3R5bGVJbnZhbGlkYXRpb25Gb3JJZENo YW5nZSk6CisgICAgICAgIChXZWJDb3JlKToKKyAgICAgICAgKFdlYkNvcmU6OkVsZW1lbnQ6OmF0 dHJpYnV0ZUNoYW5nZWQpOgorICAgIAorICAgICAgICAgICAgQ2xlYW4gdXAgaWQgdGVzdGluZyB0 b28uCisKKyAgICAgICAgKFdlYkNvcmU6OmNoZWNrTmVlZHNTdHlsZUludmFsaWRhdGlvbkZvckNs YXNzQ2hhbmdlKToKKyAgICAgICAgCisgICAgICAgICAgICBVc2UgYml0IHZlY3RvciBmb3IgbWFy a2luZyBzZWVuIHZhbHVlcy4gQXZvaWRzIGFsbG9jYXRpb25zIGFuZCByZWZmaW5nLgorCisgICAg ICAgIChXZWJDb3JlOjpFbGVtZW50OjpjbGFzc0F0dHJpYnV0ZUNoYW5nZWQpOgorICAgICAgICAK KyAgICAgICAgICAgIERvbid0IHRlc3QgaWYgc3R5bGUgaXMgYWxyZWFkeSBpbnZhbGlkLgorCiAy MDEyLTExLTAxICBQYXZlbCBGZWxkbWFuICA8cGZlbGRtYW5AY2hyb21pdW0ub3JnPgogCiAgICAg ICAgIFdlYiBJbnNwZWN0b3I6IFVwZGF0ZSBDb2RlTWlycm9yIHRvIHYzCkluZGV4OiBTb3VyY2Uv V2ViQ29yZS9jc3MvU3R5bGVSZXNvbHZlci5jcHAKPT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291cmNlL1dlYkNv cmUvY3NzL1N0eWxlUmVzb2x2ZXIuY3BwCShyZXZpc2lvbiAxMzI5NDEpCisrKyBTb3VyY2UvV2Vi Q29yZS9jc3MvU3R5bGVSZXNvbHZlci5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTQyNDksMjUgKzQy NDksNiBAQCBDb2xvciBTdHlsZVJlc29sdmVyOjpjb2xvckZyb21QcmltaXRpdmVWCiAgICAgfQog fQogCi1ib29sIFN0eWxlUmVzb2x2ZXI6Omhhc1NlbGVjdG9yRm9yQXR0cmlidXRlKGNvbnN0IEF0 b21pY1N0cmluZyAmYXR0cm5hbWUpIGNvbnN0Ci17Ci0gICAgcmV0dXJuIG1fZmVhdHVyZXMuYXR0 cnNJblJ1bGVzLmNvbnRhaW5zKGF0dHJuYW1lLmltcGwoKSk7Ci19Ci0KLWJvb2wgU3R5bGVSZXNv bHZlcjo6aGFzU2VsZWN0b3JGb3JDbGFzcyhjb25zdCBBdG9taWNTdHJpbmcmIGNsYXNzVmFsdWUp IGNvbnN0Ci17Ci0gICAgaWYgKGNsYXNzVmFsdWUuaXNFbXB0eSgpKQotICAgICAgICByZXR1cm4g ZmFsc2U7Ci0gICAgcmV0dXJuIG1fZmVhdHVyZXMuY2xhc3Nlc0luUnVsZXMuY29udGFpbnMoY2xh c3NWYWx1ZS5pbXBsKCkpOwotfQotCi1ib29sIFN0eWxlUmVzb2x2ZXI6Omhhc1NlbGVjdG9yRm9y SWQoY29uc3QgQXRvbWljU3RyaW5nJiBpZFZhbHVlKSBjb25zdAotewotICAgIGlmIChpZFZhbHVl LmlzRW1wdHkoKSkKLSAgICAgICAgcmV0dXJuIGZhbHNlOwotICAgIHJldHVybiBtX2ZlYXR1cmVz Lmlkc0luUnVsZXMuY29udGFpbnMoaWRWYWx1ZS5pbXBsKCkpOwotfQotCiB2b2lkIFN0eWxlUmVz b2x2ZXI6OmFkZFZpZXdwb3J0RGVwZW5kZW50TWVkaWFRdWVyeVJlc3VsdChjb25zdCBNZWRpYVF1 ZXJ5RXhwKiBleHByLCBib29sIHJlc3VsdCkKIHsKICAgICBtX3ZpZXdwb3J0RGVwZW5kZW50TWVk aWFRdWVyeVJlc3VsdHMuYXBwZW5kKGFkb3B0UHRyKG5ldyBNZWRpYVF1ZXJ5UmVzdWx0KCpleHBy LCByZXN1bHQpKSk7CkluZGV4OiBTb3VyY2UvV2ViQ29yZS9jc3MvU3R5bGVSZXNvbHZlci5oCj09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT0KLS0tIFNvdXJjZS9XZWJDb3JlL2Nzcy9TdHlsZVJlc29sdmVyLmgJKHJldmlzaW9u IDEzMjk0MSkKKysrIFNvdXJjZS9XZWJDb3JlL2Nzcy9TdHlsZVJlc29sdmVyLmgJKHdvcmtpbmcg Y29weSkKQEAgLTUxOCw2ICs1MTgsMjQgQEAgcHJpdmF0ZToKICAgICBmcmllbmQgYm9vbCBvcGVy YXRvciE9KGNvbnN0IE1hdGNoUmFuZ2VzJiwgY29uc3QgTWF0Y2hSYW5nZXMmKTsKIH07CiAKK2lu bGluZSBib29sIFN0eWxlUmVzb2x2ZXI6Omhhc1NlbGVjdG9yRm9yQXR0cmlidXRlKGNvbnN0IEF0 b21pY1N0cmluZyAmYXR0cmlidXRlTmFtZSkgY29uc3QKK3sKKyAgICBBU1NFUlQoIWF0dHJpYnV0 ZU5hbWUuaXNFbXB0eSgpKTsKKyAgICByZXR1cm4gbV9mZWF0dXJlcy5hdHRyc0luUnVsZXMuY29u dGFpbnMoYXR0cmlidXRlTmFtZS5pbXBsKCkpOworfQorCitpbmxpbmUgYm9vbCBTdHlsZVJlc29s dmVyOjpoYXNTZWxlY3RvckZvckNsYXNzKGNvbnN0IEF0b21pY1N0cmluZyYgY2xhc3NWYWx1ZSkg Y29uc3QKK3sKKyAgICBBU1NFUlQoIWNsYXNzVmFsdWUuaXNFbXB0eSgpKTsKKyAgICByZXR1cm4g bV9mZWF0dXJlcy5jbGFzc2VzSW5SdWxlcy5jb250YWlucyhjbGFzc1ZhbHVlLmltcGwoKSk7Cit9 CisKK2lubGluZSBib29sIFN0eWxlUmVzb2x2ZXI6Omhhc1NlbGVjdG9yRm9ySWQoY29uc3QgQXRv bWljU3RyaW5nJiBpZFZhbHVlKSBjb25zdAoreworICAgIEFTU0VSVCghaWRWYWx1ZS5pc0VtcHR5 KCkpOworICAgIHJldHVybiBtX2ZlYXR1cmVzLmlkc0luUnVsZXMuY29udGFpbnMoaWRWYWx1ZS5p bXBsKCkpOworfQorCiB9IC8vIG5hbWVzcGFjZSBXZWJDb3JlCiAKICNlbmRpZiAvLyBTdHlsZVJl c29sdmVyX2gKSW5kZXg6IFNvdXJjZS9XZWJDb3JlL2RvbS9FbGVtZW50LmNwcAo9PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 Ci0tLSBTb3VyY2UvV2ViQ29yZS9kb20vRWxlbWVudC5jcHAJKHJldmlzaW9uIDEzMzAyMSkKKysr IFNvdXJjZS9XZWJDb3JlL2RvbS9FbGVtZW50LmNwcAkod29ya2luZyBjb3B5KQpAQCAtNzcsNiAr NzcsNyBAQAogI2luY2x1ZGUgIlhNTE5TTmFtZXMuaCIKICNpbmNsdWRlICJYTUxOYW1lcy5oIgog I2luY2x1ZGUgImh0bWxlZGl0aW5nLmgiCisjaW5jbHVkZSA8d3RmL0JpdFZlY3Rvci5oPgogI2lu Y2x1ZGUgPHd0Zi90ZXh0L0NTdHJpbmcuaD4KIAogI2lmIEVOQUJMRShTVkcpCkBAIC03MDIsMzIg KzcwMyw0NSBAQCBzdGF0aWMgaW5saW5lIEF0b21pY1N0cmluZyBtYWtlSWRGb3JTdHlsCiAgICAg cmV0dXJuIHZhbHVlOwogfQogCitzdGF0aWMgYm9vbCBjaGVja05lZWRzU3R5bGVJbnZhbGlkYXRp b25Gb3JJZENoYW5nZShjb25zdCBBdG9taWNTdHJpbmcmIG9sZElkLCBjb25zdCBBdG9taWNTdHJp bmcmIG5ld0lkLCBTdHlsZVJlc29sdmVyKiBzdHlsZVJlc29sdmVyKQoreworICAgIEFTU0VSVChu ZXdJZCAhPSBvbGRJZCk7CisgICAgaWYgKCFvbGRJZC5pc0VtcHR5KCkgJiYgc3R5bGVSZXNvbHZl ci0+aGFzU2VsZWN0b3JGb3JJZChvbGRJZCkpCisgICAgICAgIHJldHVybiB0cnVlOworICAgIGlm ICghbmV3SWQuaXNFbXB0eSgpICYmIHN0eWxlUmVzb2x2ZXItPmhhc1NlbGVjdG9yRm9ySWQobmV3 SWQpKQorICAgICAgICByZXR1cm4gdHJ1ZTsKKyAgICByZXR1cm4gZmFsc2U7Cit9CisKIHZvaWQg RWxlbWVudDo6YXR0cmlidXRlQ2hhbmdlZChjb25zdCBRdWFsaWZpZWROYW1lJiBuYW1lLCBjb25z dCBBdG9taWNTdHJpbmcmIG5ld1ZhbHVlKQogewogICAgIHBhcnNlQXR0cmlidXRlKEF0dHJpYnV0 ZShuYW1lLCBuZXdWYWx1ZSkpOwogCiAgICAgZG9jdW1lbnQoKS0+aW5jRE9NVHJlZVZlcnNpb24o KTsKIAorICAgIFN0eWxlUmVzb2x2ZXIqIHN0eWxlUmVzb2x2ZXIgPSBkb2N1bWVudCgpLT5zdHls ZVJlc29sdmVySWZFeGlzdHMoKTsKKyAgICBib29sIHRlc3RTaG91bGRJbnZhbGlkYXRlU3R5bGUg PSBhdHRhY2hlZCgpICYmIHN0eWxlUmVzb2x2ZXIgJiYgc3R5bGVDaGFuZ2VUeXBlKCkgPCBGdWxs U3R5bGVDaGFuZ2U7CisgICAgYm9vbCBzaG91bGRJbnZhbGlkYXRlU3R5bGUgPSBmYWxzZTsKKwog ICAgIGlmIChpc0lkQXR0cmlidXRlTmFtZShuYW1lKSkgewogICAgICAgICBBdG9taWNTdHJpbmcg b2xkSWQgPSBhdHRyaWJ1dGVEYXRhKCktPmlkRm9yU3R5bGVSZXNvbHV0aW9uKCk7CiAgICAgICAg IEF0b21pY1N0cmluZyBuZXdJZCA9IG1ha2VJZEZvclN0eWxlUmVzb2x1dGlvbihuZXdWYWx1ZSwg ZG9jdW1lbnQoKS0+aW5RdWlya3NNb2RlKCkpOwogICAgICAgICBpZiAobmV3SWQgIT0gb2xkSWQp IHsKICAgICAgICAgICAgIGF0dHJpYnV0ZURhdGEoKS0+c2V0SWRGb3JTdHlsZVJlc29sdXRpb24o bmV3SWQpOwotICAgICAgICAgICAgU3R5bGVSZXNvbHZlciogc3R5bGVSZXNvbHZlciA9IGRvY3Vt ZW50KCktPnN0eWxlUmVzb2x2ZXJJZkV4aXN0cygpOwotICAgICAgICAgICAgaWYgKGF0dGFjaGVk KCkgJiYgKCFzdHlsZVJlc29sdmVyIHx8IChzdHlsZVJlc29sdmVyLT5oYXNTZWxlY3RvckZvcklk KG5ld0lkKSB8fCBzdHlsZVJlc29sdmVyLT5oYXNTZWxlY3RvckZvcklkKG9sZElkKSkpKQotICAg ICAgICAgICAgICAgIHNldE5lZWRzU3R5bGVSZWNhbGMoKTsKKyAgICAgICAgICAgIHNob3VsZElu dmFsaWRhdGVTdHlsZSA9IHRlc3RTaG91bGRJbnZhbGlkYXRlU3R5bGUgJiYgY2hlY2tOZWVkc1N0 eWxlSW52YWxpZGF0aW9uRm9ySWRDaGFuZ2Uob2xkSWQsIG5ld0lkLCBzdHlsZVJlc29sdmVyKTsK KyAgICAgICAgICAgIGlmIChzaG91bGRJbnZhbGlkYXRlU3R5bGUpCisgICAgICAgICAgICAgICAg dGVzdFNob3VsZEludmFsaWRhdGVTdHlsZSA9IGZhbHNlOwogICAgICAgICB9CiAgICAgfSBlbHNl IGlmIChuYW1lID09IEhUTUxOYW1lczo6bmFtZUF0dHIpCiAgICAgICAgIHNldEhhc05hbWUoIW5l d1ZhbHVlLmlzTnVsbCgpKTsKIAotICAgIGlmICghbmVlZHNTdHlsZVJlY2FsYygpICYmIGRvY3Vt ZW50KCktPmF0dGFjaGVkKCkpIHsKLSAgICAgICAgU3R5bGVSZXNvbHZlciogc3R5bGVSZXNvbHZl ciA9IGRvY3VtZW50KCktPnN0eWxlUmVzb2x2ZXJJZkV4aXN0cygpOwotICAgICAgICBpZiAoIXN0 eWxlUmVzb2x2ZXIgfHwgc3R5bGVSZXNvbHZlci0+aGFzU2VsZWN0b3JGb3JBdHRyaWJ1dGUobmFt ZS5sb2NhbE5hbWUoKSkpCi0gICAgICAgICAgICBzZXROZWVkc1N0eWxlUmVjYWxjKCk7Ci0gICAg fQorICAgIHNob3VsZEludmFsaWRhdGVTdHlsZSA9IHRlc3RTaG91bGRJbnZhbGlkYXRlU3R5bGUg JiYgc3R5bGVSZXNvbHZlci0+aGFzU2VsZWN0b3JGb3JBdHRyaWJ1dGUobmFtZS5sb2NhbE5hbWUo KSk7CiAKICAgICBpbnZhbGlkYXRlTm9kZUxpc3RDYWNoZXNJbkFuY2VzdG9ycygmbmFtZSwgdGhp cyk7CiAKKyAgICBpZiAoc2hvdWxkSW52YWxpZGF0ZVN0eWxlKQorICAgICAgICBzZXROZWVkc1N0 eWxlUmVjYWxjKCk7CisKICAgICBpZiAoQVhPYmplY3RDYWNoZTo6YWNjZXNzaWJpbGl0eUVuYWJs ZWQoKSkKICAgICAgICAgZG9jdW1lbnQoKS0+YXhPYmplY3RDYWNoZSgpLT5oYW5kbGVBdHRyaWJ1 dGVDaGFuZ2VkKG5hbWUsIHRoaXMpOwogfQpAQCAtNzY1LDQ5ICs3NzksNzIgQEAgc3RhdGljIGlu bGluZSBib29sIGNsYXNzU3RyaW5nSGFzQ2xhc3NOYQogICAgIHJldHVybiBjbGFzc1N0cmluZ0hh c0NsYXNzTmFtZShuZXdDbGFzc1N0cmluZy5jaGFyYWN0ZXJzMTYoKSwgbGVuZ3RoKTsKIH0KIAot c3RhdGljIHZvaWQgY29sbGVjdEFkZGVkQW5kUmVtb3ZlZENsYXNzZXMoVmVjdG9yPEF0b21pY1N0 cmluZywgND4mIGFkZGVkQW5kUmVtb3ZlZENsYXNzZXMsIFNwYWNlU3BsaXRTdHJpbmcmIG9sZENs YXNzZXMsIGNvbnN0IFNwYWNlU3BsaXRTdHJpbmcmIG5ld0NsYXNzZXMpCitzdGF0aWMgYm9vbCBj aGVja05lZWRzU3R5bGVJbnZhbGlkYXRpb25Gb3JDbGFzc0NoYW5nZShjb25zdCBTcGFjZVNwbGl0 U3RyaW5nJiBjaGFuZ2VkQ2xhc3NlcywgU3R5bGVSZXNvbHZlciogc3R5bGVSZXNvbHZlcikKK3sK KyAgICB1bnNpZ25lZCBjaGFuZ2VkU2l6ZSA9IGNoYW5nZWRDbGFzc2VzLnNpemUoKTsKKyAgICBm b3IgKHVuc2lnbmVkIGkgPSAwOyBpIDwgY2hhbmdlZFNpemU7ICsraSkgeworICAgICAgICBpZiAo c3R5bGVSZXNvbHZlci0+aGFzU2VsZWN0b3JGb3JDbGFzcyhjaGFuZ2VkQ2xhc3Nlc1tpXSkpCisg ICAgICAgICAgICByZXR1cm4gdHJ1ZTsKKyAgICB9CisgICAgcmV0dXJuIGZhbHNlOworfQorCitz dGF0aWMgYm9vbCBjaGVja05lZWRzU3R5bGVJbnZhbGlkYXRpb25Gb3JDbGFzc0NoYW5nZShjb25z dCBTcGFjZVNwbGl0U3RyaW5nJiBvbGRDbGFzc2VzLCBjb25zdCBTcGFjZVNwbGl0U3RyaW5nJiBu ZXdDbGFzc2VzLCBTdHlsZVJlc29sdmVyKiBzdHlsZVJlc29sdmVyKQogewotICAgIGZvciAodW5z aWduZWQgaSA9IDA7IGkgPCBuZXdDbGFzc2VzLnNpemUoKTsgKytpKSB7Ci0gICAgICAgIGlmIChv bGRDbGFzc2VzLnJlbW92ZShuZXdDbGFzc2VzW2ldKSkKKyAgICB1bnNpZ25lZCBvbGRTaXplID0g b2xkQ2xhc3Nlcy5zaXplKCk7CisgICAgaWYgKCFvbGRTaXplKQorICAgICAgICByZXR1cm4gY2hl Y2tOZWVkc1N0eWxlSW52YWxpZGF0aW9uRm9yQ2xhc3NDaGFuZ2UobmV3Q2xhc3Nlcywgc3R5bGVS ZXNvbHZlcik7CisgICAgQml0VmVjdG9yIHJlbWFpbmluZ0NsYXNzQml0czsKKyAgICByZW1haW5p bmdDbGFzc0JpdHMuZW5zdXJlU2l6ZShvbGRDbGFzc2VzLnNpemUoKSk7CisgICAgLy8gQ2xhc3Mg dmVjdG9ycyB0ZW5kIHRvIGJlIHZlcnkgc2hvcnQuIFRoaXMgaXMgZmFzdGVyIHRoYW4gdXNpbmcg YSBoYXNoIHRhYmxlLgorICAgIHVuc2lnbmVkIG5ld1NpemUgPSBuZXdDbGFzc2VzLnNpemUoKTsK KyAgICBmb3IgKHVuc2lnbmVkIGkgPSAwOyBpIDwgbmV3U2l6ZTsgKytpKSB7CisgICAgICAgIGZv ciAodW5zaWduZWQgaiA9IDA7IGogPCBvbGRTaXplOyArK2opIHsKKyAgICAgICAgICAgIGlmIChu ZXdDbGFzc2VzW2ldID09IG9sZENsYXNzZXNbal0pIHsKKyAgICAgICAgICAgICAgICByZW1haW5p bmdDbGFzc0JpdHMucXVpY2tTZXQoaik7CisgICAgICAgICAgICAgICAgY29udGludWU7CisgICAg ICAgICAgICB9CisgICAgICAgIH0KKyAgICAgICAgaWYgKHN0eWxlUmVzb2x2ZXItPmhhc1NlbGVj dG9yRm9yQ2xhc3MobmV3Q2xhc3Nlc1tpXSkpCisgICAgICAgICAgICByZXR1cm4gdHJ1ZTsKKyAg ICB9CisgICAgZm9yICh1bnNpZ25lZCBpID0gMDsgaSA8IG9sZFNpemU7ICsraSkgeworICAgICAg ICAvLyBJZiB0aGUgYml0IGlzIG5vdCBzZXQgdGhlIHRoZSBjb3JyZXNwb25kaW5nIGNsYXNzIGhh cyBiZWVuIHJlbW92ZWQuCisgICAgICAgIGlmIChyZW1haW5pbmdDbGFzc0JpdHMucXVpY2tHZXQo aSkpCiAgICAgICAgICAgICBjb250aW51ZTsKLSAgICAgICAgYWRkZWRBbmRSZW1vdmVkQ2xhc3Nl cy5hcHBlbmQobmV3Q2xhc3Nlc1tpXSk7CisgICAgICAgIGlmIChzdHlsZVJlc29sdmVyLT5oYXNT ZWxlY3RvckZvckNsYXNzKG9sZENsYXNzZXNbaV0pKQorICAgICAgICAgICAgcmV0dXJuIHRydWU7 CiAgICAgfQotICAgIGZvciAodW5zaWduZWQgaSA9IDA7IGkgPCBvbGRDbGFzc2VzLnNpemUoKTsg KytpKQotICAgICAgICBhZGRlZEFuZFJlbW92ZWRDbGFzc2VzLmFwcGVuZChvbGRDbGFzc2VzW2ld KTsKKyAgICByZXR1cm4gZmFsc2U7CiB9CiAKIHZvaWQgRWxlbWVudDo6Y2xhc3NBdHRyaWJ1dGVD aGFuZ2VkKGNvbnN0IEF0b21pY1N0cmluZyYgbmV3Q2xhc3NTdHJpbmcpCiB7Ci0gICAgYm9vbCBz aG91bGRJbnZhbGlkYXRlU3R5bGUgPSBhdHRhY2hlZCgpICYmIGRvY3VtZW50KCktPnN0eWxlUmVz b2x2ZXJJZkV4aXN0cygpOwotICAgIFZlY3RvcjxBdG9taWNTdHJpbmcsIDQ+IGFkZGVkQW5kUmVt b3ZlZENsYXNzZXM7CisgICAgU3R5bGVSZXNvbHZlciogc3R5bGVSZXNvbHZlciA9IGRvY3VtZW50 KCktPnN0eWxlUmVzb2x2ZXJJZkV4aXN0cygpOworICAgIGJvb2wgdGVzdFNob3VsZEludmFsaWRh dGVTdHlsZSA9IGF0dGFjaGVkKCkgJiYgc3R5bGVSZXNvbHZlciAmJiBzdHlsZUNoYW5nZVR5cGUo KSA8IEZ1bGxTdHlsZUNoYW5nZTsKKyAgICBib29sIHNob3VsZEludmFsaWRhdGVTdHlsZSA9IGZh bHNlOwogCiAgICAgaWYgKGNsYXNzU3RyaW5nSGFzQ2xhc3NOYW1lKG5ld0NsYXNzU3RyaW5nKSkg ewogICAgICAgICBjb25zdCBFbGVtZW50QXR0cmlidXRlRGF0YSogYXR0cmlidXRlRGF0YSA9IGVu c3VyZUF0dHJpYnV0ZURhdGEoKTsKICAgICAgICAgY29uc3QgYm9vbCBzaG91bGRGb2xkQ2FzZSA9 IGRvY3VtZW50KCktPmluUXVpcmtzTW9kZSgpOwotICAgICAgICBTcGFjZVNwbGl0U3RyaW5nIG9s ZENsYXNzZXMgPSBhdHRyaWJ1dGVEYXRhLT5jbGFzc05hbWVzKCk7CisgICAgICAgIGNvbnN0IFNw YWNlU3BsaXRTdHJpbmcgb2xkQ2xhc3NlcyA9IGF0dHJpYnV0ZURhdGEtPmNsYXNzTmFtZXMoKTsK KwogICAgICAgICBhdHRyaWJ1dGVEYXRhLT5zZXRDbGFzcyhuZXdDbGFzc1N0cmluZywgc2hvdWxk Rm9sZENhc2UpOwotICAgICAgICBpZiAoc2hvdWxkSW52YWxpZGF0ZVN0eWxlKQotICAgICAgICAg ICAgY29sbGVjdEFkZGVkQW5kUmVtb3ZlZENsYXNzZXMoYWRkZWRBbmRSZW1vdmVkQ2xhc3Nlcywg b2xkQ2xhc3NlcywgYXR0cmlidXRlRGF0YS0+Y2xhc3NOYW1lcygpKTsKKworICAgICAgICBjb25z dCBTcGFjZVNwbGl0U3RyaW5nJiBuZXdDbGFzc2VzID0gYXR0cmlidXRlRGF0YS0+Y2xhc3NOYW1l cygpOworICAgICAgICBzaG91bGRJbnZhbGlkYXRlU3R5bGUgPSB0ZXN0U2hvdWxkSW52YWxpZGF0 ZVN0eWxlICYmIGNoZWNrTmVlZHNTdHlsZUludmFsaWRhdGlvbkZvckNsYXNzQ2hhbmdlKG9sZENs YXNzZXMsIG5ld0NsYXNzZXMsIHN0eWxlUmVzb2x2ZXIpOwogICAgIH0gZWxzZSBpZiAoY29uc3Qg RWxlbWVudEF0dHJpYnV0ZURhdGEqIGF0dHJpYnV0ZURhdGEgPSB0aGlzLT5hdHRyaWJ1dGVEYXRh KCkpIHsKLSAgICAgICAgaWYgKHNob3VsZEludmFsaWRhdGVTdHlsZSkgewotICAgICAgICAgICAg Y29uc3QgU3BhY2VTcGxpdFN0cmluZyYgb2xkQ2xhc3NlcyA9IGF0dHJpYnV0ZURhdGEtPmNsYXNz TmFtZXMoKTsKLSAgICAgICAgICAgIGZvciAodW5zaWduZWQgaSA9IDA7IGkgPCBvbGRDbGFzc2Vz LnNpemUoKTsgKytpKQotICAgICAgICAgICAgICAgIGFkZGVkQW5kUmVtb3ZlZENsYXNzZXMuYXBw ZW5kKG9sZENsYXNzZXNbaV0pOwotICAgICAgICB9CisgICAgICAgIGNvbnN0IFNwYWNlU3BsaXRT dHJpbmcmIG9sZENsYXNzZXMgPSBhdHRyaWJ1dGVEYXRhLT5jbGFzc05hbWVzKCk7CisgICAgICAg IHNob3VsZEludmFsaWRhdGVTdHlsZSA9IHRlc3RTaG91bGRJbnZhbGlkYXRlU3R5bGUgJiYgY2hl Y2tOZWVkc1N0eWxlSW52YWxpZGF0aW9uRm9yQ2xhc3NDaGFuZ2Uob2xkQ2xhc3Nlcywgc3R5bGVS ZXNvbHZlcik7CisKICAgICAgICAgYXR0cmlidXRlRGF0YS0+Y2xlYXJDbGFzcygpOwogICAgIH0K IAogICAgIGlmIChET01Ub2tlbkxpc3QqIGNsYXNzTGlzdCA9IG9wdGlvbmFsQ2xhc3NMaXN0KCkp CiAgICAgICAgIHN0YXRpY19jYXN0PENsYXNzTGlzdCo+KGNsYXNzTGlzdCktPnJlc2V0KG5ld0Ns YXNzU3RyaW5nKTsKIAotICAgIGlmICghc2hvdWxkSW52YWxpZGF0ZVN0eWxlKQotICAgICAgICBy ZXR1cm47Ci0gICAgZm9yICh1bnNpZ25lZCBpID0gMDsgaSA8IGFkZGVkQW5kUmVtb3ZlZENsYXNz ZXMuc2l6ZSgpOyArK2kpIHsKLSAgICAgICAgaWYgKGRvY3VtZW50KCktPnN0eWxlUmVzb2x2ZXJJ ZkV4aXN0cygpLT5oYXNTZWxlY3RvckZvckNsYXNzKGFkZGVkQW5kUmVtb3ZlZENsYXNzZXNbaV0p KSB7Ci0gICAgICAgICAgICBzZXROZWVkc1N0eWxlUmVjYWxjKCk7Ci0gICAgICAgICAgICByZXR1 cm47Ci0gICAgICAgIH0KLSAgICB9CisgICAgaWYgKHNob3VsZEludmFsaWRhdGVTdHlsZSkKKyAg ICAgICAgc2V0TmVlZHNTdHlsZVJlY2FsYygpOwogfQogCiAvLyBSZXR1cm5zIHRydWUgaXMgdGhl IGdpdmVuIGF0dHJpYnV0ZSBpcyBhbiBldmVudCBoYW5kbGVyLgo=