158678 2016-06-12 17:07:50 -0700 The sampling profiler should further protect itself against certain forms of sampling bias that arise due to the sampling interval being in sync with some other system process 2016-06-13 19:27:55 -0700 1 1 1 Unclassified WebKit JavaScriptCore WebKit Local Build Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 saam saam bburg benjamin commit-queue fpizlo ggaren gskachkov joepeck keith_miller mark.lam msaboff naserid13 oliver sukolsak ysuzuki oldest_to_newest 1201736 0 saam 2016-06-12 17:07:50 -0700 This problem was first brought to my attention after reading this paper. See section 6.2: http://plv.colorado.edu/papers/mytkowicz-pldi10.pdf It seems somewhat rare that this will happen in practice, and it's unlikely to happen many times in a row after restarting the profiler, but it's good practice to defend against this form of bias. 1201857 1 281176 saam 2016-06-13 09:31:15 -0700 Created attachment 281176 patch 1201858 2 commit-queue 2016-06-13 09:32:55 -0700 Attachment 281176 did not pass style-queue: ERROR: Source/JavaScriptCore/runtime/SamplingProfiler.cpp:52: Alphabetical sorting problem. [build/include_order] [4] Total errors found: 1 in 2 files If any of these errors are false positives, please file a bug against check-webkit-style. 1201897 3 281176 benjamin 2016-06-13 12:03:54 -0700 Comment on attachment 281176 patch View in context: https://bugs.webkit.org/attachment.cgi?id=281176&action=review > Source/JavaScriptCore/runtime/SamplingProfiler.cpp:237 > + std::chrono::microseconds randomFluctuation = std::chrono::microseconds(static_cast<uint64_t>(randomNumber() * (static_cast<double>(m_timingInterval.count()) * 0.20l))); Is this right? I understand a distribution around "t" but randomNumber() is always positive, m_timingInterval.count() is always positive. Shouldn't the distribution be spread on both sides of t? This line would also benefit from mentioning the paper in a comment. 1201914 4 281176 saam 2016-06-13 12:45:32 -0700 Comment on attachment 281176 patch View in context: https://bugs.webkit.org/attachment.cgi?id=281176&action=review >> Source/JavaScriptCore/runtime/SamplingProfiler.cpp:237 >> + std::chrono::microseconds randomFluctuation = std::chrono::microseconds(static_cast<uint64_t>(randomNumber() * (static_cast<double>(m_timingInterval.count()) * 0.20l))); > > Is this right? > > I understand a distribution around "t" but randomNumber() is always positive, m_timingInterval.count() is always positive. > Shouldn't the distribution be spread on both sides of t? > > This line would also benefit from mentioning the paper in a comment. Yeah, obviously. I wasn't thinking. This is definitely wrong. I shall fix and update with a link. Thanks for catching this 1202038 5 281221 saam 2016-06-13 18:40:09 -0700 Created attachment 281221 patch 1202040 6 commit-queue 2016-06-13 18:42:03 -0700 Attachment 281221 did not pass style-queue: ERROR: Source/JavaScriptCore/runtime/SamplingProfiler.cpp:52: Alphabetical sorting problem. [build/include_order] [4] Total errors found: 1 in 2 files If any of these errors are false positives, please file a bug against check-webkit-style. 1202042 7 281221 benjamin 2016-06-13 18:46:01 -0700 Comment on attachment 281221 patch View in context: https://bugs.webkit.org/attachment.cgi?id=281221&action=review > Source/JavaScriptCore/runtime/SamplingProfiler.cpp:241 > + double randomSignedNumber = (randomNumber() * 2.0l) - 1.0l; // A random number between [-1, 1). What's the .01? Why not "(randomNumber() * 2.) - 1"? > Source/JavaScriptCore/runtime/SamplingProfiler.cpp:242 > + std::chrono::microseconds randomFluctuation = std::chrono::microseconds(static_cast<uint64_t>(randomSignedNumber * static_cast<double>(m_timingInterval.count()) * 0.20l)); What's that "* 0.20l" factor here? 1202043 8 281221 saam 2016-06-13 18:48:48 -0700 Comment on attachment 281221 patch View in context: https://bugs.webkit.org/attachment.cgi?id=281221&action=review >> Source/JavaScriptCore/runtime/SamplingProfiler.cpp:241 >> + double randomSignedNumber = (randomNumber() * 2.0l) - 1.0l; // A random number between [-1, 1). > > What's the .01? > Why not "(randomNumber() * 2.) - 1"? it's an 'L' not a 1. For a double literal. I'll just switch to 2.0 and 1.0 >> Source/JavaScriptCore/runtime/SamplingProfiler.cpp:242 >> + std::chrono::microseconds randomFluctuation = std::chrono::microseconds(static_cast<uint64_t>(randomSignedNumber * static_cast<double>(m_timingInterval.count()) * 0.20l)); > > What's that "* 0.20l" factor here? We add +-(0.2t) where t is our sampling interval. We're just choosing some range between [-t, t] 1202046 9 281223 saam 2016-06-13 18:50:02 -0700 Created attachment 281223 patch 1202051 10 saam 2016-06-13 18:56:58 -0700 (In reply to comment #8) > Comment on attachment 281221 [details] > patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=281221&action=review > > >> Source/JavaScriptCore/runtime/SamplingProfiler.cpp:241 > >> + double randomSignedNumber = (randomNumber() * 2.0l) - 1.0l; // A random number between [-1, 1). > > > > What's the .01? > > Why not "(randomNumber() * 2.) - 1"? > > it's an 'L' not a 1. > For a double literal. I'll just switch to 2.0 and 1.0 > > >> Source/JavaScriptCore/runtime/SamplingProfiler.cpp:242 > >> + std::chrono::microseconds randomFluctuation = std::chrono::microseconds(static_cast<uint64_t>(randomSignedNumber * static_cast<double>(m_timingInterval.count()) * 0.20l)); > > > > What's that "* 0.20l" factor here? > > We add +-(0.2t) where t is our sampling interval. > We're just choosing some range between [-t, t] To elaborate on this, we don't do the full [-t, t] because we want to defend against longer sample times for short running programs. Even though short running programs are hard to run under a sampling profiler, we don't want to hurt such uses. We choose 0.2 because a random 20% of our default (1ms) range is enough to add randomness to our sample times so we don't end up in sync with some other system process. 1202054 11 281223 commit-queue 2016-06-13 19:27:50 -0700 Comment on attachment 281223 patch Clearing flags on attachment: 281223 Committed r202021: <http://trac.webkit.org/changeset/202021> 1202055 12 commit-queue 2016-06-13 19:27:55 -0700 All reviewed patches have been landed. Closing bug. 281176 2016-06-13 09:31:15 -0700 2016-06-13 18:40:09 -0700 patch c-backup.diff text/plain 2944 saam SW5kZXg6IFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291 cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkocmV2aXNpb24gMjAxOTg1KQorKysgU291cmNl L0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDMwIEBA CisyMDE2LTA2LTEyICBTYWFtIEJhcmF0aSAgPHNiYXJhdGlAYXBwbGUuY29tPgorCisgICAgICAg IFRoZSBzYW1wbGluZyBwcm9maWxlciBzaG91bGQgZnVydGhlciBwcm90ZWN0IGl0c2VsZiBhZ2Fp bnN0IGNlcnRhaW4gZm9ybXMgb2Ygc2FtcGxpbmcgYmlhcyB0aGF0IGFyaXNlIGR1ZSB0byB0aGUg c2FtcGxpbmcgaW50ZXJ2YWwgYmVpbmcgaW4gc3luYyB3aXRoIHNvbWUgb3RoZXIgc3lzdGVtIHBy b2Nlc3MKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE1 ODY3OAorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIEkg Zmlyc3QgYmVjYW1lIGF3YXJlIG9mIHRoaXMgcHJvYmxlbSB3aGVuIEkgcmVhZCB0aGlzIHBhcGVy OgorICAgICAgICBodHRwOi8vcGx2LmNvbG9yYWRvLmVkdS9wYXBlcnMvbXl0a293aWN6LXBsZGkx MC5wZGYKKworICAgICAgICBUbyBwcm92aWRlIGJhY2tncm91bmQgZm9yIHRoaXMgY2hhbmdlLCBJ J2xsIHF1b3RlIGEgcGFyYWdyYXBoCisgICAgICAgIGZyb20gc2VjdGlvbiA2LjI6CisgICAgICAg ICJPbmUgc3RhdGljYWxseSBzb3VuZCBtZXRob2QgZm9yIGNvbGxlY3RpbmcgcmFuZG9tIHNhbXBs ZXMgaXMgdG8gY29sbGVjdCBhCisgICAgICAgIHNhbXBsZSBhdCBldmVyeSB0ICsgciBtaWxsaXNl Y29uZHMsIHdoZXJlIHQgaXMgdGhlIGRlc2lyZWQgc2FtcGxpbmcgaW50ZXJ2YWwKKyAgICAgICAg YW5kIHIgaXMgYSByYW5kb20gbnVtYmVyIGJldHdlZW4g4oiSdCBhbmQgdC4gT25lIG1pZ2h0IHRo aW5rIHRoYXQgc2FtcGxpbmcgZXZlcnkKKyAgICAgICAgdCBzZWNvbmRzIGlzIGVub3VnaCAoaS5l LiwgZHJvcCB0aGUgciBjb21wb25lbnQpIGJ1dCBpdCBpcyBub3Q6IHNwZWNpZmljYWxseSwKKyAg ICAgICAgaWYgYSBwcm9maWxlciBzYW1wbGVzIGV2ZXJ5IHQgc2Vjb25kcywgdGhlIHNhbXBsaW5n IHJhdGUgd291bGQgYmUgc3luY2hyb25pemVkCisgICAgICAgIHdpdGggYW55IHByb2dyYW0gb3Ig c3lzdGVtIGFjdGl2aXR5IHRoYXQgb2NjdXJzIGF0IHJlZ3VsYXIgdGltZSBpbnRlcnZhbHMgWzE3 XS4KKyAgICAgICAgRm9yIGV4YW1wbGUsIGlmIHRoZSB0aHJlYWQgc2NoZWR1bGVyIHN3aXRjaGVz IGJldHdlZW4gdGhyZWFkcyBldmVyeSAxMG1zIGFuZCBvdXIKKyAgICAgICAgc2FtcGxpbmcgaW50 ZXJ2YWwgd2FzIGFsc28gMTBtcywgdGhlbiB3ZSBtYXkgYWx3YXlzIHRha2Ugc2FtcGxlcyBpbW1l ZGlhdGVseSBhZnRlcgorICAgICAgICBhIHRocmVhZCBzd2l0Y2guIEJlY2F1c2UgcGVyZm9ybWFu Y2UgaXMgb2Z0ZW4gZGlmZmVyZW50IGltbWVkaWF0ZWx5IGFmdGVyIGEgdGhyZWFkCisgICAgICAg IHN3aXRjaCB0aGFuIGF0IG90aGVyIHBvaW50cyAoZS5nLiwgZHVlIHRvIGNhY2hlIGFuZCBUTEIg d2FybS11cCBlZmZlY3RzKSB3ZSB3b3VsZAorICAgICAgICBnZXQgYmlhc2VkIGRhdGEuIFRoZSBy YW5kb20gY29tcG9uZW50LCByLCBndWFyZHMgYWdhaW5zdCBzdWNoIHNpdHVhdGlvbnMuIgorCisg ICAgICAgICogcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcDoKKyAgICAgICAgKEpTQzo6U2Ft cGxpbmdQcm9maWxlcjo6dGltZXJMb29wKToKKwogMjAxNi0wNi0xMSAgTWFyayBMYW0gIDxtYXJr LmxhbUBhcHBsZS5jb20+CiAKICAgICAgICAgTWluaW1pemUgdGhlIGFtb3VudCBvZiBtZW1jcHkg ZG9uZSBmb3IgYWxsb2NhdGluZyBFcnJvciBzdGFja3MuCkluZGV4OiBTb3VyY2UvSmF2YVNjcmlw dENvcmUvcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcAo9PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2Uv SmF2YVNjcmlwdENvcmUvcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcAkocmV2aXNpb24gMjAx OTg1KQorKysgU291cmNlL0phdmFTY3JpcHRDb3JlL3J1bnRpbWUvU2FtcGxpbmdQcm9maWxlci5j cHAJKHdvcmtpbmcgY29weSkKQEAgLTQ5LDYgKzQ5LDcgQEAKICNpbmNsdWRlICJWTUVudHJ5U2Nv cGUuaCIKICNpbmNsdWRlIDx3dGYvSGFzaFNldC5oPgogI2luY2x1ZGUgPHd0Zi9SZWZQdHIuaD4K KyNpbmNsdWRlIDx3dGYvUmFuZG9tTnVtYmVyLmg+CiAKIG5hbWVzcGFjZSBKU0MgewogCkBAIC0y MzMsNyArMjM0LDggQEAgdm9pZCBTYW1wbGluZ1Byb2ZpbGVyOjp0aW1lckxvb3AoKQogICAgICAg ICAgICAgbV9sYXN0VGltZSA9IG1fc3RvcHdhdGNoLT5lbGFwc2VkVGltZSgpOwogICAgICAgICB9 CiAKLSAgICAgICAgc3RkOjp0aGlzX3RocmVhZDo6c2xlZXBfZm9yKG1fdGltaW5nSW50ZXJ2YWwg LSBzdGQ6Om1pbihtX3RpbWluZ0ludGVydmFsLCBzdGFja1RyYWNlUHJvY2Vzc2luZ1RpbWUpKTsK KyAgICAgICAgc3RkOjpjaHJvbm86Om1pY3Jvc2Vjb25kcyByYW5kb21GbHVjdHVhdGlvbiA9IHN0 ZDo6Y2hyb25vOjptaWNyb3NlY29uZHMoc3RhdGljX2Nhc3Q8dWludDY0X3Q+KHJhbmRvbU51bWJl cigpICogKHN0YXRpY19jYXN0PGRvdWJsZT4obV90aW1pbmdJbnRlcnZhbC5jb3VudCgpKSAqIDAu MjBsKSkpOworICAgICAgICBzdGQ6OnRoaXNfdGhyZWFkOjpzbGVlcF9mb3IobV90aW1pbmdJbnRl cnZhbCAtIHN0ZDo6bWluKG1fdGltaW5nSW50ZXJ2YWwsIHN0YWNrVHJhY2VQcm9jZXNzaW5nVGlt ZSkgKyByYW5kb21GbHVjdHVhdGlvbik7CiAgICAgfQogfQogCg== 281221 2016-06-13 18:40:09 -0700 2016-06-13 18:50:02 -0700 patch c-backup.diff text/plain 3363 saam SW5kZXg6IFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291 cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkocmV2aXNpb24gMjAxOTg1KQorKysgU291cmNl L0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDMwIEBA CisyMDE2LTA2LTEyICBTYWFtIEJhcmF0aSAgPHNiYXJhdGlAYXBwbGUuY29tPgorCisgICAgICAg IFRoZSBzYW1wbGluZyBwcm9maWxlciBzaG91bGQgZnVydGhlciBwcm90ZWN0IGl0c2VsZiBhZ2Fp bnN0IGNlcnRhaW4gZm9ybXMgb2Ygc2FtcGxpbmcgYmlhcyB0aGF0IGFyaXNlIGR1ZSB0byB0aGUg c2FtcGxpbmcgaW50ZXJ2YWwgYmVpbmcgaW4gc3luYyB3aXRoIHNvbWUgb3RoZXIgc3lzdGVtIHBy b2Nlc3MKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE1 ODY3OAorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIEkg Zmlyc3QgYmVjYW1lIGF3YXJlIG9mIHRoaXMgcHJvYmxlbSB3aGVuIEkgcmVhZCB0aGlzIHBhcGVy OgorICAgICAgICBodHRwOi8vcGx2LmNvbG9yYWRvLmVkdS9wYXBlcnMvbXl0a293aWN6LXBsZGkx MC5wZGYKKworICAgICAgICBUbyBwcm92aWRlIGJhY2tncm91bmQgZm9yIHRoaXMgY2hhbmdlLCBJ J2xsIHF1b3RlIGEgcGFyYWdyYXBoCisgICAgICAgIGZyb20gc2VjdGlvbiA2LjI6CisgICAgICAg ICJPbmUgc3RhdGljYWxseSBzb3VuZCBtZXRob2QgZm9yIGNvbGxlY3RpbmcgcmFuZG9tIHNhbXBs ZXMgaXMgdG8gY29sbGVjdCBhCisgICAgICAgIHNhbXBsZSBhdCBldmVyeSB0ICsgciBtaWxsaXNl Y29uZHMsIHdoZXJlIHQgaXMgdGhlIGRlc2lyZWQgc2FtcGxpbmcgaW50ZXJ2YWwKKyAgICAgICAg YW5kIHIgaXMgYSByYW5kb20gbnVtYmVyIGJldHdlZW4g4oiSdCBhbmQgdC4gT25lIG1pZ2h0IHRo aW5rIHRoYXQgc2FtcGxpbmcgZXZlcnkKKyAgICAgICAgdCBzZWNvbmRzIGlzIGVub3VnaCAoaS5l LiwgZHJvcCB0aGUgciBjb21wb25lbnQpIGJ1dCBpdCBpcyBub3Q6IHNwZWNpZmljYWxseSwKKyAg ICAgICAgaWYgYSBwcm9maWxlciBzYW1wbGVzIGV2ZXJ5IHQgc2Vjb25kcywgdGhlIHNhbXBsaW5n IHJhdGUgd291bGQgYmUgc3luY2hyb25pemVkCisgICAgICAgIHdpdGggYW55IHByb2dyYW0gb3Ig c3lzdGVtIGFjdGl2aXR5IHRoYXQgb2NjdXJzIGF0IHJlZ3VsYXIgdGltZSBpbnRlcnZhbHMgWzE3 XS4KKyAgICAgICAgRm9yIGV4YW1wbGUsIGlmIHRoZSB0aHJlYWQgc2NoZWR1bGVyIHN3aXRjaGVz IGJldHdlZW4gdGhyZWFkcyBldmVyeSAxMG1zIGFuZCBvdXIKKyAgICAgICAgc2FtcGxpbmcgaW50 ZXJ2YWwgd2FzIGFsc28gMTBtcywgdGhlbiB3ZSBtYXkgYWx3YXlzIHRha2Ugc2FtcGxlcyBpbW1l ZGlhdGVseSBhZnRlcgorICAgICAgICBhIHRocmVhZCBzd2l0Y2guIEJlY2F1c2UgcGVyZm9ybWFu Y2UgaXMgb2Z0ZW4gZGlmZmVyZW50IGltbWVkaWF0ZWx5IGFmdGVyIGEgdGhyZWFkCisgICAgICAg IHN3aXRjaCB0aGFuIGF0IG90aGVyIHBvaW50cyAoZS5nLiwgZHVlIHRvIGNhY2hlIGFuZCBUTEIg d2FybS11cCBlZmZlY3RzKSB3ZSB3b3VsZAorICAgICAgICBnZXQgYmlhc2VkIGRhdGEuIFRoZSBy YW5kb20gY29tcG9uZW50LCByLCBndWFyZHMgYWdhaW5zdCBzdWNoIHNpdHVhdGlvbnMuIgorCisg ICAgICAgICogcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcDoKKyAgICAgICAgKEpTQzo6U2Ft cGxpbmdQcm9maWxlcjo6dGltZXJMb29wKToKKwogMjAxNi0wNi0xMSAgTWFyayBMYW0gIDxtYXJr LmxhbUBhcHBsZS5jb20+CiAKICAgICAgICAgTWluaW1pemUgdGhlIGFtb3VudCBvZiBtZW1jcHkg ZG9uZSBmb3IgYWxsb2NhdGluZyBFcnJvciBzdGFja3MuCkluZGV4OiBTb3VyY2UvSmF2YVNjcmlw dENvcmUvcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcAo9PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2Uv SmF2YVNjcmlwdENvcmUvcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcAkocmV2aXNpb24gMjAx OTg1KQorKysgU291cmNlL0phdmFTY3JpcHRDb3JlL3J1bnRpbWUvU2FtcGxpbmdQcm9maWxlci5j cHAJKHdvcmtpbmcgY29weSkKQEAgLTQ5LDYgKzQ5LDcgQEAKICNpbmNsdWRlICJWTUVudHJ5U2Nv cGUuaCIKICNpbmNsdWRlIDx3dGYvSGFzaFNldC5oPgogI2luY2x1ZGUgPHd0Zi9SZWZQdHIuaD4K KyNpbmNsdWRlIDx3dGYvUmFuZG9tTnVtYmVyLmg+CiAKIG5hbWVzcGFjZSBKU0MgewogCkBAIC0y MzMsNyArMjM0LDEzIEBAIHZvaWQgU2FtcGxpbmdQcm9maWxlcjo6dGltZXJMb29wKCkKICAgICAg ICAgICAgIG1fbGFzdFRpbWUgPSBtX3N0b3B3YXRjaC0+ZWxhcHNlZFRpbWUoKTsKICAgICAgICAg fQogCi0gICAgICAgIHN0ZDo6dGhpc190aHJlYWQ6OnNsZWVwX2ZvcihtX3RpbWluZ0ludGVydmFs IC0gc3RkOjptaW4obV90aW1pbmdJbnRlcnZhbCwgc3RhY2tUcmFjZVByb2Nlc3NpbmdUaW1lKSk7 CisgICAgICAgIC8vIFJlYWQgc2VjdGlvbiA2LjIgb2YgdGhpcyBwYXBlciBmb3IgbW9yZSBlbGFi b3JhdGlvbiBvZiB3aHkgd2UgYWRkIGEgcmFuZG9tCisgICAgICAgIC8vIGZsdWN0dWF0aW9uIGhl cmUuIFRoZSBtYWluIGlkZWEgaXMgdG8gcHJldmVudCBvdXIgdGltZXIgZnJvbSBiZWluZyBpbiBz eW5jCisgICAgICAgIC8vIHdpdGggc29tZSBzeXN0ZW0gcHJvY2VzcyBzdWNoIGFzIGEgc2NoZWR1 bGVkIGNvbnRleHQgc3dpdGNoLgorICAgICAgICAvLyBodHRwOi8vcGx2LmNvbG9yYWRvLmVkdS9w YXBlcnMvbXl0a293aWN6LXBsZGkxMC5wZGYKKyAgICAgICAgZG91YmxlIHJhbmRvbVNpZ25lZE51 bWJlciA9IChyYW5kb21OdW1iZXIoKSAqIDIuMGwpIC0gMS4wbDsgLy8gQSByYW5kb20gbnVtYmVy IGJldHdlZW4gWy0xLCAxKS4KKyAgICAgICAgc3RkOjpjaHJvbm86Om1pY3Jvc2Vjb25kcyByYW5k b21GbHVjdHVhdGlvbiA9IHN0ZDo6Y2hyb25vOjptaWNyb3NlY29uZHMoc3RhdGljX2Nhc3Q8dWlu dDY0X3Q+KHJhbmRvbVNpZ25lZE51bWJlciAqIHN0YXRpY19jYXN0PGRvdWJsZT4obV90aW1pbmdJ bnRlcnZhbC5jb3VudCgpKSAqIDAuMjBsKSk7CisgICAgICAgIHN0ZDo6dGhpc190aHJlYWQ6OnNs ZWVwX2ZvcihtX3RpbWluZ0ludGVydmFsIC0gc3RkOjptaW4obV90aW1pbmdJbnRlcnZhbCwgc3Rh Y2tUcmFjZVByb2Nlc3NpbmdUaW1lKSArIHJhbmRvbUZsdWN0dWF0aW9uKTsKICAgICB9CiB9CiAK 281223 2016-06-13 18:50:02 -0700 2016-06-13 19:27:50 -0700 patch c-backup.diff text/plain 3376 saam SW5kZXg6IFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291 cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkocmV2aXNpb24gMjAxOTg1KQorKysgU291cmNl L0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDMwIEBA CisyMDE2LTA2LTEyICBTYWFtIEJhcmF0aSAgPHNiYXJhdGlAYXBwbGUuY29tPgorCisgICAgICAg IFRoZSBzYW1wbGluZyBwcm9maWxlciBzaG91bGQgZnVydGhlciBwcm90ZWN0IGl0c2VsZiBhZ2Fp bnN0IGNlcnRhaW4gZm9ybXMgb2Ygc2FtcGxpbmcgYmlhcyB0aGF0IGFyaXNlIGR1ZSB0byB0aGUg c2FtcGxpbmcgaW50ZXJ2YWwgYmVpbmcgaW4gc3luYyB3aXRoIHNvbWUgb3RoZXIgc3lzdGVtIHBy b2Nlc3MKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTE1 ODY3OAorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIEkg Zmlyc3QgYmVjYW1lIGF3YXJlIG9mIHRoaXMgcHJvYmxlbSB3aGVuIEkgcmVhZCB0aGlzIHBhcGVy OgorICAgICAgICBodHRwOi8vcGx2LmNvbG9yYWRvLmVkdS9wYXBlcnMvbXl0a293aWN6LXBsZGkx MC5wZGYKKworICAgICAgICBUbyBwcm92aWRlIGJhY2tncm91bmQgZm9yIHRoaXMgY2hhbmdlLCBJ J2xsIHF1b3RlIGEgcGFyYWdyYXBoCisgICAgICAgIGZyb20gc2VjdGlvbiA2LjI6CisgICAgICAg ICJPbmUgc3RhdGljYWxseSBzb3VuZCBtZXRob2QgZm9yIGNvbGxlY3RpbmcgcmFuZG9tIHNhbXBs ZXMgaXMgdG8gY29sbGVjdCBhCisgICAgICAgIHNhbXBsZSBhdCBldmVyeSB0ICsgciBtaWxsaXNl Y29uZHMsIHdoZXJlIHQgaXMgdGhlIGRlc2lyZWQgc2FtcGxpbmcgaW50ZXJ2YWwKKyAgICAgICAg YW5kIHIgaXMgYSByYW5kb20gbnVtYmVyIGJldHdlZW4g4oiSdCBhbmQgdC4gT25lIG1pZ2h0IHRo aW5rIHRoYXQgc2FtcGxpbmcgZXZlcnkKKyAgICAgICAgdCBzZWNvbmRzIGlzIGVub3VnaCAoaS5l LiwgZHJvcCB0aGUgciBjb21wb25lbnQpIGJ1dCBpdCBpcyBub3Q6IHNwZWNpZmljYWxseSwKKyAg ICAgICAgaWYgYSBwcm9maWxlciBzYW1wbGVzIGV2ZXJ5IHQgc2Vjb25kcywgdGhlIHNhbXBsaW5n IHJhdGUgd291bGQgYmUgc3luY2hyb25pemVkCisgICAgICAgIHdpdGggYW55IHByb2dyYW0gb3Ig c3lzdGVtIGFjdGl2aXR5IHRoYXQgb2NjdXJzIGF0IHJlZ3VsYXIgdGltZSBpbnRlcnZhbHMgWzE3 XS4KKyAgICAgICAgRm9yIGV4YW1wbGUsIGlmIHRoZSB0aHJlYWQgc2NoZWR1bGVyIHN3aXRjaGVz IGJldHdlZW4gdGhyZWFkcyBldmVyeSAxMG1zIGFuZCBvdXIKKyAgICAgICAgc2FtcGxpbmcgaW50 ZXJ2YWwgd2FzIGFsc28gMTBtcywgdGhlbiB3ZSBtYXkgYWx3YXlzIHRha2Ugc2FtcGxlcyBpbW1l ZGlhdGVseSBhZnRlcgorICAgICAgICBhIHRocmVhZCBzd2l0Y2guIEJlY2F1c2UgcGVyZm9ybWFu Y2UgaXMgb2Z0ZW4gZGlmZmVyZW50IGltbWVkaWF0ZWx5IGFmdGVyIGEgdGhyZWFkCisgICAgICAg IHN3aXRjaCB0aGFuIGF0IG90aGVyIHBvaW50cyAoZS5nLiwgZHVlIHRvIGNhY2hlIGFuZCBUTEIg d2FybS11cCBlZmZlY3RzKSB3ZSB3b3VsZAorICAgICAgICBnZXQgYmlhc2VkIGRhdGEuIFRoZSBy YW5kb20gY29tcG9uZW50LCByLCBndWFyZHMgYWdhaW5zdCBzdWNoIHNpdHVhdGlvbnMuIgorCisg ICAgICAgICogcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcDoKKyAgICAgICAgKEpTQzo6U2Ft cGxpbmdQcm9maWxlcjo6dGltZXJMb29wKToKKwogMjAxNi0wNi0xMSAgTWFyayBMYW0gIDxtYXJr LmxhbUBhcHBsZS5jb20+CiAKICAgICAgICAgTWluaW1pemUgdGhlIGFtb3VudCBvZiBtZW1jcHkg ZG9uZSBmb3IgYWxsb2NhdGluZyBFcnJvciBzdGFja3MuCkluZGV4OiBTb3VyY2UvSmF2YVNjcmlw dENvcmUvcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcAo9PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2Uv SmF2YVNjcmlwdENvcmUvcnVudGltZS9TYW1wbGluZ1Byb2ZpbGVyLmNwcAkocmV2aXNpb24gMjAx OTg1KQorKysgU291cmNlL0phdmFTY3JpcHRDb3JlL3J1bnRpbWUvU2FtcGxpbmdQcm9maWxlci5j cHAJKHdvcmtpbmcgY29weSkKQEAgLTQ4LDYgKzQ4LDcgQEAKICNpbmNsdWRlICJWTS5oIgogI2lu Y2x1ZGUgIlZNRW50cnlTY29wZS5oIgogI2luY2x1ZGUgPHd0Zi9IYXNoU2V0Lmg+CisjaW5jbHVk ZSA8d3RmL1JhbmRvbU51bWJlci5oPgogI2luY2x1ZGUgPHd0Zi9SZWZQdHIuaD4KIAogbmFtZXNw YWNlIEpTQyB7CkBAIC0yMzMsNyArMjM0LDEzIEBAIHZvaWQgU2FtcGxpbmdQcm9maWxlcjo6dGlt ZXJMb29wKCkKICAgICAgICAgICAgIG1fbGFzdFRpbWUgPSBtX3N0b3B3YXRjaC0+ZWxhcHNlZFRp bWUoKTsKICAgICAgICAgfQogCi0gICAgICAgIHN0ZDo6dGhpc190aHJlYWQ6OnNsZWVwX2Zvciht X3RpbWluZ0ludGVydmFsIC0gc3RkOjptaW4obV90aW1pbmdJbnRlcnZhbCwgc3RhY2tUcmFjZVBy b2Nlc3NpbmdUaW1lKSk7CisgICAgICAgIC8vIFJlYWQgc2VjdGlvbiA2LjIgb2YgdGhpcyBwYXBl ciBmb3IgbW9yZSBlbGFib3JhdGlvbiBvZiB3aHkgd2UgYWRkIGEgcmFuZG9tCisgICAgICAgIC8v IGZsdWN0dWF0aW9uIGhlcmUuIFRoZSBtYWluIGlkZWEgaXMgdG8gcHJldmVudCBvdXIgdGltZXIg ZnJvbSBiZWluZyBpbiBzeW5jCisgICAgICAgIC8vIHdpdGggc29tZSBzeXN0ZW0gcHJvY2VzcyBz dWNoIGFzIGEgc2NoZWR1bGVkIGNvbnRleHQgc3dpdGNoLgorICAgICAgICAvLyBodHRwOi8vcGx2 LmNvbG9yYWRvLmVkdS9wYXBlcnMvbXl0a293aWN6LXBsZGkxMC5wZGYKKyAgICAgICAgZG91Ymxl IHJhbmRvbVNpZ25lZE51bWJlciA9IChyYW5kb21OdW1iZXIoKSAqIDIuMCkgLSAxLjA7IC8vIEEg cmFuZG9tIG51bWJlciBiZXR3ZWVuIFstMSwgMSkuCisgICAgICAgIHN0ZDo6Y2hyb25vOjptaWNy b3NlY29uZHMgcmFuZG9tRmx1Y3R1YXRpb24gPSBzdGQ6OmNocm9ubzo6bWljcm9zZWNvbmRzKHN0 YXRpY19jYXN0PHVpbnQ2NF90PihyYW5kb21TaWduZWROdW1iZXIgKiBzdGF0aWNfY2FzdDxkb3Vi bGU+KG1fdGltaW5nSW50ZXJ2YWwuY291bnQoKSkgKiAwLjIwbCkpOworICAgICAgICBzdGQ6OnRo aXNfdGhyZWFkOjpzbGVlcF9mb3IobV90aW1pbmdJbnRlcnZhbCAtIHN0ZDo6bWluKG1fdGltaW5n SW50ZXJ2YWwsIHN0YWNrVHJhY2VQcm9jZXNzaW5nVGltZSkgKyByYW5kb21GbHVjdHVhdGlvbik7 CiAgICAgfQogfQogCg==