71427 2011-11-02 18:13:30 -0700 Inlined uses of the global object should use the right global object 2011-11-02 19:59:13 -0700 1 1 1 Unclassified WebKit JavaScriptCore 528+ (Nightly build) All All RESOLVED FIXED P2 Normal --- 1 fpizlo webkit-unassigned oldest_to_newest 495123 0 fpizlo 2011-11-02 18:13:30 -0700 The DFG backend was assuming that uses of the global object can use the global object associated with the machine code block. But they should be the ones associated with the (potentially inlined) function instead. 495124 1 113413 fpizlo 2011-11-02 18:14:44 -0700 Created attachment 113413 the patch 495155 2 113413 oliver 2011-11-02 19:30:06 -0700 Comment on attachment 113413 the patch View in context: https://bugs.webkit.org/attachment.cgi?id=113413&action=review r=me, but i don't believe that fix me is correct. > Source/JavaScriptCore/dfg/DFGJITCompiler.h:439 > + // FIXME: if we ever inline based on executable not function, this code will need to change. I don't believe that this is correct -- all functions for a given executable must by definition share the same global object. 495156 3 fpizlo 2011-11-02 19:31:01 -0700 (In reply to comment #2) > (From update of attachment 113413 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=113413&action=review > > r=me, but i don't believe that fix me is correct. > > > Source/JavaScriptCore/dfg/DFGJITCompiler.h:439 > > + // FIXME: if we ever inline based on executable not function, this code will need to change. > > I don't believe that this is correct -- all functions for a given executable must by definition share the same global object. If we inline based on executable, the inlineCallFrame->callee field will be 0, and there will be another field that tells you how to get the callee from the register file. That's what I meant by the FIXME. 495164 4 fpizlo 2011-11-02 19:52:23 -0700 Looks performance neutral and all tests pass. Benchmark report for SunSpider, V8, and Kraken. VMs tested: "TipOfTree" at /Volumes/Data/pizlo/tertiary/OpenSource/WebKitBuild/Release/jsc "FixInlineGlobal" at /Volumes/Data/pizlo/secondary/OpenSource/WebKitBuild/Release/jsc Collected 12 samples per benchmark/VM, with 4 VM invocations per benchmark. Emitted a call to gc() between sample measurements. Used 1 benchmark iteration per VM invocation for warm-up. Used the jsc-specific preciseTime() function to get microsecond-level timing. Reporting benchmark execution times with 95% confidence intervals in milliseconds. TipOfTree FixInlineGlobal SunSpider: 3d-cube 7.9055+-0.0250 7.9030+-0.0245 3d-morph 8.4010+-0.0327 8.3968+-0.0273 3d-raytrace 8.2108+-0.0663 8.1345+-0.0946 access-binary-trees 1.7229+-0.0395 1.6849+-0.0058 might be 1.0226x faster access-fannkuch 7.8228+-0.0070 ? 7.8424+-0.0260 ? access-nbody 4.5250+-0.0182 ? 4.5274+-0.0092 ? access-nsieve 3.1817+-0.0122 ? 3.2215+-0.0327 ? might be 1.0125x slower bitops-3bit-bits-in-byte 1.2934+-0.0156 ? 1.2998+-0.0149 ? bitops-bits-in-byte 4.9984+-0.0076 4.9842+-0.0120 bitops-bitwise-and 3.4312+-0.0617 ? 3.4583+-0.0728 ? bitops-nsieve-bits 5.6541+-0.0357 ? 5.6871+-0.0369 ? controlflow-recursive 2.3309+-0.0122 ? 2.3598+-0.0281 ? might be 1.0124x slower crypto-aes 7.7440+-0.0641 7.7323+-0.0451 crypto-md5 2.8935+-0.0283 2.8830+-0.0122 crypto-sha1 2.6501+-0.0400 2.6337+-0.0210 date-format-tofte 10.8381+-0.1213 10.7141+-0.0460 might be 1.0116x faster date-format-xparb 10.0780+-0.1998 ^ 9.7163+-0.1056 ^ definitely 1.0372x faster math-cordic 7.2948+-0.0635 7.2656+-0.0419 math-partial-sums 10.5843+-0.0257 10.5822+-0.0750 math-spectral-norm 2.9087+-0.0334 2.8803+-0.0069 regexp-dna 13.4231+-0.1595 ? 13.4949+-0.1883 ? string-base64 4.3142+-0.0267 4.2876+-0.0140 string-fasta 7.1573+-0.0114 ^ 7.1321+-0.0136 ^ definitely 1.0035x faster string-tagcloud 13.2550+-0.1296 ? 13.2607+-0.1008 ? string-unpack-code 23.0323+-0.1087 22.9692+-0.1619 string-validate-input 5.6430+-0.0265 ? 5.6622+-0.0404 ? <arithmetic> * 6.9729+-0.0207 6.9505+-0.0238 <geometric> 5.6266+-0.0158 5.6126+-0.0150 <harmonic> 4.4490+-0.0168 4.4406+-0.0118 TipOfTree FixInlineGlobal V8: crypto 81.1349+-0.2988 80.8813+-0.2027 deltablue 182.6440+-0.9418 182.6097+-1.9133 earley-boyer 111.4420+-0.2287 111.3792+-0.4176 raytrace 69.1373+-0.3317 ! 70.0228+-0.5185 ! definitely 1.0128x slower regexp 125.9373+-0.5356 125.8627+-0.1749 richards 144.1570+-0.5759 ? 145.0371+-0.3479 ? splay 90.5705+-0.9146 90.4614+-1.1537 <arithmetic> 115.0033+-0.1811 ? 115.1792+-0.2915 ? <geometric> * 109.4122+-0.1859 ? 109.6156+-0.2958 ? <harmonic> 104.2286+-0.2096 ? 104.4777+-0.3322 ? TipOfTree FixInlineGlobal Kraken: ai-astar 818.0541+-10.8419 ? 827.8603+-0.5665 ? might be 1.0120x slower audio-beat-detection 209.9056+-1.0301 209.7902+-0.8383 audio-dft 262.9032+-1.8822 260.7352+-2.4678 audio-fft 136.9000+-0.5795 ? 137.1937+-0.7243 ? audio-oscillator 317.5338+-30.9262 290.2611+-1.1621 might be 1.0940x faster imaging-darkroom 338.8633+-5.6019 338.4315+-5.8906 imaging-desaturate 240.9064+-0.0799 ? 240.9227+-0.1222 ? imaging-gaussian-blur 621.3248+-0.9010 621.1927+-0.3395 json-parse-financial 70.4761+-0.1563 ? 70.5841+-0.1556 ? json-stringify-tinderbox 79.2334+-0.2014 ? 79.5120+-0.6848 ? stanford-crypto-aes 116.6959+-0.7287 116.5538+-0.6824 stanford-crypto-ccm 114.7526+-0.5635 ! 115.8885+-0.5538 ! definitely 1.0099x slower stanford-crypto-pbkdf2 234.7142+-0.5188 ? 235.8879+-1.0809 ? stanford-crypto-sha256-iterative 97.5235+-0.2353 ? 97.6753+-0.1444 ? <arithmetic> * 261.4134+-2.5696 260.1778+-0.3637 <geometric> 200.4197+-1.3682 199.6099+-0.2592 <harmonic> 159.7703+-0.5276 159.6421+-0.2686 TipOfTree FixInlineGlobal All benchmarks: <arithmetic> 98.8533+-0.7571 98.4991+-0.1172 <geometric> 25.3745+-0.0676 25.3164+-0.0442 <harmonic> 7.8349+-0.0288 7.8206+-0.0204 TipOfTree FixInlineGlobal Geomean of preferred means: <scaled-result> 58.4237+-0.1967 58.3067+-0.1016 495165 5 fpizlo 2011-11-02 19:59:13 -0700 Landed in http://trac.webkit.org/changeset/99132 113413 2011-11-02 18:14:44 -0700 2011-11-02 19:30:05 -0700 the patch fixglobal_patch_1.diff text/plain 10366 fpizlo SW5kZXg6IFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291 cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkocmV2aXNpb24gOTkxMjcpCisrKyBTb3VyY2Uv SmF2YVNjcmlwdENvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBAIC0xLDMgKzEsMTUgQEAK KzIwMTEtMTEtMDIgIEZpbGlwIFBpemxvICA8ZnBpemxvQGFwcGxlLmNvbT4KKworICAgICAgICBJ bmxpbmVkIHVzZXMgb2YgdGhlIGdsb2JhbCBvYmplY3Qgc2hvdWxkIHVzZSB0aGUgcmlnaHQgZ2xv YmFsIG9iamVjdAorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/ aWQ9NzE0MjcKKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAg ICAqIGRmZy9ERkdKSVRDb21waWxlci5oOgorICAgICAgICAoSlNDOjpERkc6OkpJVENvbXBpbGVy OjpnbG9iYWxPYmplY3RGb3IpOgorICAgICAgICAqIGRmZy9ERkdTcGVjdWxhdGl2ZUpJVDY0LmNw cDoKKyAgICAgICAgKEpTQzo6REZHOjpTcGVjdWxhdGl2ZUpJVDo6Y29tcGlsZSk6CisKIDIwMTEt MTEtMDIgIE1hcmsgSGFobmVuYmVyZyAgPG1oYWhuZW5iZXJnQGFwcGxlLmNvbT4KIAogICAgICAg ICBEZS12aXJ0dWFsaXplIEpTT2JqZWN0OjpnZXRPd25Qcm9wZXJ0eU5hbWVzCkluZGV4OiBTb3Vy Y2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR0pJVENvbXBpbGVyLmgKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291 cmNlL0phdmFTY3JpcHRDb3JlL2RmZy9ERkdKSVRDb21waWxlci5oCShyZXZpc2lvbiA5OTExMikK KysrIFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9kZmcvREZHSklUQ29tcGlsZXIuaAkod29ya2luZyBj b3B5KQpAQCAtNDMyLDYgKzQzMiwxNCBAQCBwdWJsaWM6CiAgICAgICAgIHJldHVybiBtX2dyYXBo LnZhbHVlUHJvZmlsZUZvcihub2RlSW5kZXgsIGJhc2VsaW5lQ29kZUJsb2NrRm9yKG1fZ3JhcGhb bm9kZUluZGV4XS5jb2RlT3JpZ2luKSk7CiAgICAgfQogICAgIAorICAgIEpTR2xvYmFsT2JqZWN0 KiBnbG9iYWxPYmplY3RGb3IoQ29kZU9yaWdpbiBjb2RlT3JpZ2luKQorICAgIHsKKyAgICAgICAg aWYgKCFjb2RlT3JpZ2luLmlubGluZUNhbGxGcmFtZSkKKyAgICAgICAgICAgIHJldHVybiBjb2Rl QmxvY2soKS0+Z2xvYmFsT2JqZWN0KCk7CisgICAgICAgIC8vIEZJWE1FOiBpZiB3ZSBldmVyIGlu bGluZSBiYXNlZCBvbiBleGVjdXRhYmxlIG5vdCBmdW5jdGlvbiwgdGhpcyBjb2RlIHdpbGwgbmVl ZCB0byBjaGFuZ2UuCisgICAgICAgIHJldHVybiBjb2RlT3JpZ2luLmlubGluZUNhbGxGcmFtZS0+ Y2FsbGVlLT5zY29wZSgpLT5nbG9iYWxPYmplY3QuZ2V0KCk7CisgICAgfQorICAgIAogcHJpdmF0 ZToKICAgICAvLyBJbnRlcm5hbCBpbXBsZW1lbnRhdGlvbiB0byBjb21waWxlLgogICAgIHZvaWQg Y29tcGlsZUVudHJ5KCk7CkluZGV4OiBTb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR1NwZWN1 bGF0aXZlSklUMzJfNjQuY3BwCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9k ZmcvREZHU3BlY3VsYXRpdmVKSVQzMl82NC5jcHAJKHJldmlzaW9uIDk5MTEyKQorKysgU291cmNl L0phdmFTY3JpcHRDb3JlL2RmZy9ERkdTcGVjdWxhdGl2ZUpJVDMyXzY0LmNwcAkod29ya2luZyBj b3B5KQpAQCAtMTkwMSw3ICsxOTAxLDcgQEAgdm9pZCBTcGVjdWxhdGl2ZUpJVDo6Y29tcGlsZShO b2RlJiBub2RlKQogICAgICAgICAgICAgLy8gc2luY2UgdGhpcyBvcGVyYXRpb24gZG9lcyBub3Qg b3RoZXJ3aXNlIGdldCB0aGUgcGF5bG9hZC4KICAgICAgICAgICAgIHNwZWN1bGF0aW9uQ2hlY2so SlNWYWx1ZVJlZ3MoKSwgTm9Ob2RlLCBtX2ppdC5icmFuY2gzMihNYWNyb0Fzc2VtYmxlcjo6Tm90 RXF1YWwsIHNjcmF0Y2hHUFIsIFRydXN0ZWRJbW0zMihKU1ZhbHVlOjpOdWxsVGFnKSkpOwogICAg ICAgICAgICAgCi0gICAgICAgICAgICBtX2ppdC5tb3ZlKE1hY3JvQXNzZW1ibGVyOjpUcnVzdGVk SW1tUHRyKG1faml0LmNvZGVCbG9jaygpLT5nbG9iYWxPYmplY3QoKSksIHNjcmF0Y2hHUFIpOwor ICAgICAgICAgICAgbV9qaXQubW92ZShNYWNyb0Fzc2VtYmxlcjo6VHJ1c3RlZEltbVB0cihtX2pp dC5nbG9iYWxPYmplY3RGb3Iobm9kZS5jb2RlT3JpZ2luKSksIHNjcmF0Y2hHUFIpOwogICAgICAg ICAgICAgY2VsbFJlc3VsdChzY3JhdGNoR1BSLCBtX2NvbXBpbGVJbmRleCk7CiAgICAgICAgICAg ICBicmVhazsKICAgICAgICAgfQpAQCAtMTk5NCw3ICsxOTk0LDcgQEAgdm9pZCBTcGVjdWxhdGl2 ZUpJVDo6Y29tcGlsZShOb2RlJiBub2RlKQogICAgICAgICAKICAgICAgICAgTWFjcm9Bc3NlbWJs ZXI6Okp1bXBMaXN0IHNsb3dQYXRoOwogICAgICAgICAKLSAgICAgICAgZW1pdEFsbG9jYXRlSlNG aW5hbE9iamVjdChNYWNyb0Fzc2VtYmxlcjo6VHJ1c3RlZEltbVB0cihtX2ppdC5jb2RlQmxvY2so KS0+Z2xvYmFsT2JqZWN0KCktPmVtcHR5T2JqZWN0U3RydWN0dXJlKCkpLCByZXN1bHRHUFIsIHNj cmF0Y2hHUFIsIHNsb3dQYXRoKTsKKyAgICAgICAgZW1pdEFsbG9jYXRlSlNGaW5hbE9iamVjdChN YWNyb0Fzc2VtYmxlcjo6VHJ1c3RlZEltbVB0cihtX2ppdC5nbG9iYWxPYmplY3RGb3Iobm9kZS5j b2RlT3JpZ2luKS0+ZW1wdHlPYmplY3RTdHJ1Y3R1cmUoKSksIHJlc3VsdEdQUiwgc2NyYXRjaEdQ Uiwgc2xvd1BhdGgpOwogICAgICAgICAKICAgICAgICAgTWFjcm9Bc3NlbWJsZXI6Okp1bXAgZG9u ZSA9IG1faml0Lmp1bXAoKTsKICAgICAgICAgCkBAIC0yMjg2LDcgKzIyODYsNyBAQCB2b2lkIFNw ZWN1bGF0aXZlSklUOjpjb21waWxlKE5vZGUmIG5vZGUpCiAgICAgICAgIAogICAgICAgICBpZiAo IW1fc3RhdGUuZm9yTm9kZShub2RlLmNoaWxkMSgpKS5tX3N0cnVjdHVyZS5kb2VzTm90Q29udGFp bkFueU90aGVyVGhhbihtZXRob2RDaGVja0RhdGEuc3RydWN0dXJlKSkKICAgICAgICAgICAgIHNw ZWN1bGF0aW9uQ2hlY2soSlNWYWx1ZVJlZ3MoKSwgTm9Ob2RlLCBtX2ppdC5icmFuY2hQdHIoSklU Q29tcGlsZXI6Ok5vdEVxdWFsLCBKSVRDb21waWxlcjo6QWRkcmVzcyhiYXNlR1BSLCBKU0NlbGw6 OnN0cnVjdHVyZU9mZnNldCgpKSwgSklUQ29tcGlsZXI6OlRydXN0ZWRJbW1QdHIobWV0aG9kQ2hl Y2tEYXRhLnN0cnVjdHVyZSkpKTsKLSAgICAgICAgaWYgKG1ldGhvZENoZWNrRGF0YS5wcm90b3R5 cGUgIT0gbV9qaXQuY29kZUJsb2NrKCktPmdsb2JhbE9iamVjdCgpLT5tZXRob2RDYWxsRHVtbXko KSkgeworICAgICAgICBpZiAobWV0aG9kQ2hlY2tEYXRhLnByb3RvdHlwZSAhPSBtX2ppdC5nbG9i YWxPYmplY3RGb3Iobm9kZS5jb2RlT3JpZ2luKS0+bWV0aG9kQ2FsbER1bW15KCkpIHsKICAgICAg ICAgICAgIG1faml0Lm1vdmUoSklUQ29tcGlsZXI6OlRydXN0ZWRJbW1QdHIobWV0aG9kQ2hlY2tE YXRhLnByb3RvdHlwZS0+c3RydWN0dXJlQWRkcmVzcygpKSwgc2NyYXRjaEdQUik7CiAgICAgICAg ICAgICBzcGVjdWxhdGlvbkNoZWNrKEpTVmFsdWVSZWdzKCksIE5vTm9kZSwgbV9qaXQuYnJhbmNo UHRyKEpJVENvbXBpbGVyOjpOb3RFcXVhbCwgSklUQ29tcGlsZXI6OkFkZHJlc3Moc2NyYXRjaEdQ UiksIEpJVENvbXBpbGVyOjpUcnVzdGVkSW1tUHRyKG1ldGhvZENoZWNrRGF0YS5wcm90b3R5cGVT dHJ1Y3R1cmUpKSk7CiAgICAgICAgIH0KQEAgLTIzMzgsNyArMjMzOCw3IEBAIHZvaWQgU3BlY3Vs YXRpdmVKSVQ6OmNvbXBpbGUoTm9kZSYgbm9kZSkKICAgICAgICAgR1BSVGVtcG9yYXJ5IHJlc3Vs dCh0aGlzKTsKICAgICAgICAgR1BSVGVtcG9yYXJ5IHNjcmF0Y2godGhpcyk7CiAKLSAgICAgICAg SlNWYXJpYWJsZU9iamVjdCogZ2xvYmFsT2JqZWN0ID0gbV9qaXQuY29kZUJsb2NrKCktPmdsb2Jh bE9iamVjdCgpOworICAgICAgICBKU1ZhcmlhYmxlT2JqZWN0KiBnbG9iYWxPYmplY3QgPSBtX2pp dC5nbG9iYWxPYmplY3RGb3Iobm9kZS5jb2RlT3JpZ2luKTsKICAgICAgICAgbV9qaXQubG9hZFB0 cihjb25zdF9jYXN0PFdyaXRlQmFycmllcjxVbmtub3duPioqPihnbG9iYWxPYmplY3QtPmFkZHJl c3NPZlJlZ2lzdGVycygpKSwgcmVzdWx0LmdwcigpKTsKICAgICAgICAgbV9qaXQubG9hZDMyKEpJ VENvbXBpbGVyOjp0YWdGb3JHbG9iYWxWYXIocmVzdWx0LmdwcigpLCBub2RlLnZhck51bWJlcigp KSwgc2NyYXRjaC5ncHIoKSk7CiAgICAgICAgIG1faml0LmxvYWQzMihKSVRDb21waWxlcjo6cGF5 bG9hZEZvckdsb2JhbFZhcihyZXN1bHQuZ3ByKCksIG5vZGUudmFyTnVtYmVyKCkpLCByZXN1bHQu Z3ByKCkpOwpAQCAtMjM1NSw5ICsyMzU1LDkgQEAgdm9pZCBTcGVjdWxhdGl2ZUpJVDo6Y29tcGls ZShOb2RlJiBub2RlKQogICAgICAgICBHUFJSZWcgZ2xvYmFsT2JqZWN0UmVnID0gZ2xvYmFsT2Jq ZWN0LmdwcigpOwogICAgICAgICBHUFJSZWcgc2NyYXRjaFJlZyA9IHNjcmF0Y2guZ3ByKCk7CiAK LSAgICAgICAgbV9qaXQubW92ZShNYWNyb0Fzc2VtYmxlcjo6VHJ1c3RlZEltbVB0cihtX2ppdC5j b2RlQmxvY2soKS0+Z2xvYmFsT2JqZWN0KCkpLCBnbG9iYWxPYmplY3RSZWcpOworICAgICAgICBt X2ppdC5tb3ZlKE1hY3JvQXNzZW1ibGVyOjpUcnVzdGVkSW1tUHRyKG1faml0Lmdsb2JhbE9iamVj dEZvcihub2RlLmNvZGVPcmlnaW4pKSwgZ2xvYmFsT2JqZWN0UmVnKTsKIAotICAgICAgICB3cml0 ZUJhcnJpZXIobV9qaXQuY29kZUJsb2NrKCktPmdsb2JhbE9iamVjdCgpLCB2YWx1ZS50YWdHUFIo KSwgbm9kZS5jaGlsZDEoKSwgV3JpdGVCYXJyaWVyRm9yVmFyaWFibGVBY2Nlc3MsIHNjcmF0Y2hS ZWcpOworICAgICAgICB3cml0ZUJhcnJpZXIobV9qaXQuZ2xvYmFsT2JqZWN0Rm9yKG5vZGUuY29k ZU9yaWdpbiksIHZhbHVlLnRhZ0dQUigpLCBub2RlLmNoaWxkMSgpLCBXcml0ZUJhcnJpZXJGb3JW YXJpYWJsZUFjY2Vzcywgc2NyYXRjaFJlZyk7CiAKICAgICAgICAgbV9qaXQubG9hZFB0cihNYWNy b0Fzc2VtYmxlcjo6QWRkcmVzcyhnbG9iYWxPYmplY3RSZWcsIEpTVmFyaWFibGVPYmplY3Q6Om9m ZnNldE9mUmVnaXN0ZXJzKCkpLCBzY3JhdGNoUmVnKTsKICAgICAgICAgbV9qaXQuc3RvcmUzMih2 YWx1ZS50YWdHUFIoKSwgSklUQ29tcGlsZXI6OnRhZ0Zvckdsb2JhbFZhcihzY3JhdGNoUmVnLCBu b2RlLnZhck51bWJlcigpKSk7CkBAIC0yNDc1LDcgKzI0NzUsNyBAQCB2b2lkIFNwZWN1bGF0aXZl SklUOjpjb21waWxlKE5vZGUmIG5vZGUpCiAgICAgICAgIEdsb2JhbFJlc29sdmVJbmZvKiByZXNv bHZlSW5mb0FkZHJlc3MgPSAmKG1faml0LmNvZGVCbG9jaygpLT5nbG9iYWxSZXNvbHZlSW5mbyhk YXRhLnJlc29sdmVJbmZvSW5kZXgpKTsKIAogICAgICAgICAvLyBDaGVjayBTdHJ1Y3R1cmUgb2Yg Z2xvYmFsIG9iamVjdAotICAgICAgICBtX2ppdC5tb3ZlKEpJVENvbXBpbGVyOjpUcnVzdGVkSW1t UHRyKG1faml0LmNvZGVCbG9jaygpLT5nbG9iYWxPYmplY3QoKSksIGdsb2JhbE9iamVjdEdQUik7 CisgICAgICAgIG1faml0Lm1vdmUoSklUQ29tcGlsZXI6OlRydXN0ZWRJbW1QdHIobV9qaXQuZ2xv YmFsT2JqZWN0Rm9yKG5vZGUuY29kZU9yaWdpbikpLCBnbG9iYWxPYmplY3RHUFIpOwogICAgICAg ICBtX2ppdC5tb3ZlKEpJVENvbXBpbGVyOjpUcnVzdGVkSW1tUHRyKHJlc29sdmVJbmZvQWRkcmVz cyksIHJlc29sdmVJbmZvR1BSKTsKICAgICAgICAgbV9qaXQubG9hZFB0cihKSVRDb21waWxlcjo6 QWRkcmVzcyhyZXNvbHZlSW5mb0dQUiwgT0JKRUNUX09GRlNFVE9GKEdsb2JhbFJlc29sdmVJbmZv LCBzdHJ1Y3R1cmUpKSwgcmVzdWx0UGF5bG9hZEdQUik7CiAKSW5kZXg6IFNvdXJjZS9KYXZhU2Ny aXB0Q29yZS9kZmcvREZHU3BlY3VsYXRpdmVKSVQ2NC5jcHAKPT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291cmNl L0phdmFTY3JpcHRDb3JlL2RmZy9ERkdTcGVjdWxhdGl2ZUpJVDY0LmNwcAkocmV2aXNpb24gOTkx MTIpCisrKyBTb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR1NwZWN1bGF0aXZlSklUNjQuY3Bw CSh3b3JraW5nIGNvcHkpCkBAIC0xOTExLDcgKzE5MTEsNyBAQCB2b2lkIFNwZWN1bGF0aXZlSklU Ojpjb21waWxlKE5vZGUmIG5vZGUpCiAgICAgICAgICAgICAgICAgc3BlY3VsYXRpb25DaGVjayhK U1ZhbHVlUmVncyh0aGlzVmFsdWVHUFIpLCBub2RlLmNoaWxkMSgpLCBtX2ppdC5icmFuY2hQdHIo TWFjcm9Bc3NlbWJsZXI6Ok5vdEVxdWFsLCBzY3JhdGNoR1BSLCBNYWNyb0Fzc2VtYmxlcjo6VHJ1 c3RlZEltbVB0cihyZWludGVycHJldF9jYXN0PHZvaWQqPihWYWx1ZU51bGwpKSkpOwogICAgICAg ICAgICAgfQogICAgICAgICAgICAgCi0gICAgICAgICAgICBtX2ppdC5tb3ZlKE1hY3JvQXNzZW1i bGVyOjpUcnVzdGVkSW1tUHRyKG1faml0LmNvZGVCbG9jaygpLT5nbG9iYWxPYmplY3QoKSksIHNj cmF0Y2hHUFIpOworICAgICAgICAgICAgbV9qaXQubW92ZShNYWNyb0Fzc2VtYmxlcjo6VHJ1c3Rl ZEltbVB0cihtX2ppdC5nbG9iYWxPYmplY3RGb3Iobm9kZS5jb2RlT3JpZ2luKSksIHNjcmF0Y2hH UFIpOwogICAgICAgICAgICAgY2VsbFJlc3VsdChzY3JhdGNoR1BSLCBtX2NvbXBpbGVJbmRleCk7 CiAgICAgICAgICAgICBicmVhazsKICAgICAgICAgfQpAQCAtMjAwMyw3ICsyMDAzLDcgQEAgdm9p ZCBTcGVjdWxhdGl2ZUpJVDo6Y29tcGlsZShOb2RlJiBub2RlKQogICAgICAgICAKICAgICAgICAg TWFjcm9Bc3NlbWJsZXI6Okp1bXBMaXN0IHNsb3dQYXRoOwogICAgICAgICAKLSAgICAgICAgZW1p dEFsbG9jYXRlSlNGaW5hbE9iamVjdChNYWNyb0Fzc2VtYmxlcjo6VHJ1c3RlZEltbVB0cihtX2pp dC5jb2RlQmxvY2soKS0+Z2xvYmFsT2JqZWN0KCktPmVtcHR5T2JqZWN0U3RydWN0dXJlKCkpLCBy ZXN1bHRHUFIsIHNjcmF0Y2hHUFIsIHNsb3dQYXRoKTsKKyAgICAgICAgZW1pdEFsbG9jYXRlSlNG aW5hbE9iamVjdChNYWNyb0Fzc2VtYmxlcjo6VHJ1c3RlZEltbVB0cihtX2ppdC5nbG9iYWxPYmpl Y3RGb3Iobm9kZS5jb2RlT3JpZ2luKS0+ZW1wdHlPYmplY3RTdHJ1Y3R1cmUoKSksIHJlc3VsdEdQ Uiwgc2NyYXRjaEdQUiwgc2xvd1BhdGgpOwogICAgICAgICAKICAgICAgICAgTWFjcm9Bc3NlbWJs ZXI6Okp1bXAgZG9uZSA9IG1faml0Lmp1bXAoKTsKICAgICAgICAgCkBAIC0yMjgwLDcgKzIyODAs NyBAQCB2b2lkIFNwZWN1bGF0aXZlSklUOjpjb21waWxlKE5vZGUmIG5vZGUpCiAgICAgICAgIAog ICAgICAgICBpZiAoIW1fc3RhdGUuZm9yTm9kZShub2RlLmNoaWxkMSgpKS5tX3N0cnVjdHVyZS5k b2VzTm90Q29udGFpbkFueU90aGVyVGhhbihtZXRob2RDaGVja0RhdGEuc3RydWN0dXJlKSkKICAg ICAgICAgICAgIHNwZWN1bGF0aW9uQ2hlY2soSlNWYWx1ZVJlZ3MoKSwgTm9Ob2RlLCBtX2ppdC5i cmFuY2hQdHIoSklUQ29tcGlsZXI6Ok5vdEVxdWFsLCBKSVRDb21waWxlcjo6QWRkcmVzcyhiYXNl R1BSLCBKU0NlbGw6OnN0cnVjdHVyZU9mZnNldCgpKSwgSklUQ29tcGlsZXI6OlRydXN0ZWRJbW1Q dHIobWV0aG9kQ2hlY2tEYXRhLnN0cnVjdHVyZSkpKTsKLSAgICAgICAgaWYgKG1ldGhvZENoZWNr RGF0YS5wcm90b3R5cGUgIT0gbV9qaXQuY29kZUJsb2NrKCktPmdsb2JhbE9iamVjdCgpLT5tZXRo b2RDYWxsRHVtbXkoKSkgeworICAgICAgICBpZiAobWV0aG9kQ2hlY2tEYXRhLnByb3RvdHlwZSAh PSBtX2ppdC5nbG9iYWxPYmplY3RGb3Iobm9kZS5jb2RlT3JpZ2luKS0+bWV0aG9kQ2FsbER1bW15 KCkpIHsKICAgICAgICAgICAgIG1faml0Lm1vdmUoSklUQ29tcGlsZXI6OlRydXN0ZWRJbW1QdHIo bWV0aG9kQ2hlY2tEYXRhLnByb3RvdHlwZS0+c3RydWN0dXJlQWRkcmVzcygpKSwgc2NyYXRjaEdQ Uik7CiAgICAgICAgICAgICBzcGVjdWxhdGlvbkNoZWNrKEpTVmFsdWVSZWdzKCksIE5vTm9kZSwg bV9qaXQuYnJhbmNoUHRyKEpJVENvbXBpbGVyOjpOb3RFcXVhbCwgSklUQ29tcGlsZXI6OkFkZHJl c3Moc2NyYXRjaEdQUiksIEpJVENvbXBpbGVyOjpUcnVzdGVkSW1tUHRyKG1ldGhvZENoZWNrRGF0 YS5wcm90b3R5cGVTdHJ1Y3R1cmUpKSk7CiAgICAgICAgIH0KQEAgLTIzMjksNyArMjMyOSw3IEBA IHZvaWQgU3BlY3VsYXRpdmVKSVQ6OmNvbXBpbGUoTm9kZSYgbm9kZSkKICAgICBjYXNlIEdldEds b2JhbFZhcjogewogICAgICAgICBHUFJUZW1wb3JhcnkgcmVzdWx0KHRoaXMpOwogCi0gICAgICAg IEpTVmFyaWFibGVPYmplY3QqIGdsb2JhbE9iamVjdCA9IG1faml0LmNvZGVCbG9jaygpLT5nbG9i YWxPYmplY3QoKTsKKyAgICAgICAgSlNWYXJpYWJsZU9iamVjdCogZ2xvYmFsT2JqZWN0ID0gbV9q aXQuZ2xvYmFsT2JqZWN0Rm9yKG5vZGUuY29kZU9yaWdpbik7CiAgICAgICAgIG1faml0LmxvYWRQ dHIoZ2xvYmFsT2JqZWN0LT5hZGRyZXNzT2ZSZWdpc3RlcnMoKSwgcmVzdWx0LmdwcigpKTsKICAg ICAgICAgbV9qaXQubG9hZFB0cihKSVRDb21waWxlcjo6YWRkcmVzc0Zvckdsb2JhbFZhcihyZXN1 bHQuZ3ByKCksIG5vZGUudmFyTnVtYmVyKCkpLCByZXN1bHQuZ3ByKCkpOwogCkBAIC0yMzQ1LDkg KzIzNDUsOSBAQCB2b2lkIFNwZWN1bGF0aXZlSklUOjpjb21waWxlKE5vZGUmIG5vZGUpCiAgICAg ICAgIEdQUlJlZyBnbG9iYWxPYmplY3RSZWcgPSBnbG9iYWxPYmplY3QuZ3ByKCk7CiAgICAgICAg IEdQUlJlZyBzY3JhdGNoUmVnID0gc2NyYXRjaC5ncHIoKTsKIAotICAgICAgICBtX2ppdC5tb3Zl KE1hY3JvQXNzZW1ibGVyOjpUcnVzdGVkSW1tUHRyKG1faml0LmNvZGVCbG9jaygpLT5nbG9iYWxP YmplY3QoKSksIGdsb2JhbE9iamVjdFJlZyk7CisgICAgICAgIG1faml0Lm1vdmUoTWFjcm9Bc3Nl bWJsZXI6OlRydXN0ZWRJbW1QdHIobV9qaXQuZ2xvYmFsT2JqZWN0Rm9yKG5vZGUuY29kZU9yaWdp bikpLCBnbG9iYWxPYmplY3RSZWcpOwogCi0gICAgICAgIHdyaXRlQmFycmllcihtX2ppdC5jb2Rl QmxvY2soKS0+Z2xvYmFsT2JqZWN0KCksIHZhbHVlLmdwcigpLCBub2RlLmNoaWxkMSgpLCBXcml0 ZUJhcnJpZXJGb3JWYXJpYWJsZUFjY2Vzcywgc2NyYXRjaFJlZyk7CisgICAgICAgIHdyaXRlQmFy cmllcihtX2ppdC5nbG9iYWxPYmplY3RGb3Iobm9kZS5jb2RlT3JpZ2luKSwgdmFsdWUuZ3ByKCks IG5vZGUuY2hpbGQxKCksIFdyaXRlQmFycmllckZvclZhcmlhYmxlQWNjZXNzLCBzY3JhdGNoUmVn KTsKIAogICAgICAgICBtX2ppdC5sb2FkUHRyKE1hY3JvQXNzZW1ibGVyOjpBZGRyZXNzKGdsb2Jh bE9iamVjdFJlZywgSlNWYXJpYWJsZU9iamVjdDo6b2Zmc2V0T2ZSZWdpc3RlcnMoKSksIHNjcmF0 Y2hSZWcpOwogICAgICAgICBtX2ppdC5zdG9yZVB0cih2YWx1ZS5ncHIoKSwgSklUQ29tcGlsZXI6 OmFkZHJlc3NGb3JHbG9iYWxWYXIoc2NyYXRjaFJlZywgbm9kZS52YXJOdW1iZXIoKSkpOwpAQCAt MjQ1OSw3ICsyNDU5LDcgQEAgdm9pZCBTcGVjdWxhdGl2ZUpJVDo6Y29tcGlsZShOb2RlJiBub2Rl KQogICAgICAgICBHbG9iYWxSZXNvbHZlSW5mbyogcmVzb2x2ZUluZm9BZGRyZXNzID0gJihtX2pp dC5jb2RlQmxvY2soKS0+Z2xvYmFsUmVzb2x2ZUluZm8oZGF0YS5yZXNvbHZlSW5mb0luZGV4KSk7 CiAKICAgICAgICAgLy8gQ2hlY2sgU3RydWN0dXJlIG9mIGdsb2JhbCBvYmplY3QKLSAgICAgICAg bV9qaXQubW92ZShKSVRDb21waWxlcjo6VHJ1c3RlZEltbVB0cihtX2ppdC5jb2RlQmxvY2soKS0+ Z2xvYmFsT2JqZWN0KCkpLCBnbG9iYWxPYmplY3RHUFIpOworICAgICAgICBtX2ppdC5tb3ZlKEpJ VENvbXBpbGVyOjpUcnVzdGVkSW1tUHRyKG1faml0Lmdsb2JhbE9iamVjdEZvcihub2RlLmNvZGVP cmlnaW4pKSwgZ2xvYmFsT2JqZWN0R1BSKTsKICAgICAgICAgbV9qaXQubW92ZShKSVRDb21waWxl cjo6VHJ1c3RlZEltbVB0cihyZXNvbHZlSW5mb0FkZHJlc3MpLCByZXNvbHZlSW5mb0dQUik7CiAg ICAgICAgIG1faml0LmxvYWRQdHIoSklUQ29tcGlsZXI6OkFkZHJlc3MocmVzb2x2ZUluZm9HUFIs IE9CSkVDVF9PRkZTRVRPRihHbG9iYWxSZXNvbHZlSW5mbywgc3RydWN0dXJlKSksIHJlc3VsdEdQ Uik7CiAgICAgICAgIEpJVENvbXBpbGVyOjpKdW1wIHN0cnVjdHVyZXNNYXRjaCA9IG1faml0LmJy YW5jaFB0cihKSVRDb21waWxlcjo6RXF1YWwsIHJlc3VsdEdQUiwgSklUQ29tcGlsZXI6OkFkZHJl c3MoZ2xvYmFsT2JqZWN0R1BSLCBKU0NlbGw6OnN0cnVjdHVyZU9mZnNldCgpKSk7Cg==