147891 2015-08-11 12:17:02 -0700 DFG::ByteCodeParser shouldn't call tryGetConstantProperty() with some StructureSet if it isn't checking that the base has a structure in that StructureSet 2015-08-12 14:01:57 -0700 1 1 1 Unclassified WebKit JavaScriptCore 528+ (Nightly build) All All RESOLVED FIXED P2 Normal --- 1 fpizlo fpizlo barraclough basile_clement benjamin ggaren mark.lam mhahnenb mmirman msaboff nrotem oliver rniwa saam sam oldest_to_newest 1116635 0 fpizlo 2015-08-11 12:17:02 -0700 Patch forthcoming. 1116637 1 258735 fpizlo 2015-08-11 12:19:36 -0700 Created attachment 258735 the patch 1116660 2 258735 mark.lam 2015-08-11 13:01:38 -0700 Comment on attachment 258735 the patch View in context: https://bugs.webkit.org/attachment.cgi?id=258735&action=review r=me with suggested removal of StructureRegistrationState and m_structureRegistrationState. > Source/JavaScriptCore/ChangeLog:13 > + (JSC::DFG::Graph::assertIsRegistered): Make this always assert even before the StructureRegistrationPhase. I presume we'll deal with the fall out if this actually does assert because that would be bad, and we want to catch those cases? > Source/JavaScriptCore/dfg/DFGGraph.cpp:-1285 > - if (m_structureRegistrationState == HaveNotStartedRegistering) > - return; > - Does it make sense to still keep m_structureRegistrationState, and the StructureRegistrationState enum around? From the code, it looks like this check was their sole purpose. 1116697 3 fpizlo 2015-08-11 14:00:51 -0700 (In reply to comment #2) > Comment on attachment 258735 [details] > the patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=258735&action=review > > r=me with suggested removal of StructureRegistrationState and > m_structureRegistrationState. > > > Source/JavaScriptCore/ChangeLog:13 > > + (JSC::DFG::Graph::assertIsRegistered): Make this always assert even before the StructureRegistrationPhase. > > I presume we'll deal with the fall out if this actually does assert because > that would be bad, and we want to catch those cases? Yeah, that's the idea. Making this assert do things is what caught the bug in ByteCodeParser that this fixes, but I anticipate it will catch more bugs too. > > > Source/JavaScriptCore/dfg/DFGGraph.cpp:-1285 > > - if (m_structureRegistrationState == HaveNotStartedRegistering) > > - return; > > - > > Does it make sense to still keep m_structureRegistrationState, and the > StructureRegistrationState enum around? From the code, it looks like this > check was their sole purpose. We should get rid of all of this nonsense. This patch keeps it so that we have a quick gardening fix for now. For example, I could roll out just the change to assertIsRegistered if we get a lot of bot assertions due to this changeset. 1116702 4 fpizlo 2015-08-11 14:04:53 -0700 Landed in http://trac.webkit.org/changeset/188292 1117015 5 rniwa 2015-08-12 12:26:35 -0700 This appears to have caused 4% regression on Octane and 2% regression on JetStream. 1117017 6 fpizlo 2015-08-12 12:28:11 -0700 (In reply to comment #5) > This appears to have caused 4% regression on Octane and 2% regression on > JetStream. Looking into it. 1117031 7 fpizlo 2015-08-12 13:08:15 -0700 (In reply to comment #6) > (In reply to comment #5) > > This appears to have caused 4% regression on Octane and 2% regression on > > JetStream. > > Looking into it. Curiously the SunSpider regression doesn't repro on the command line. But the Octane one certainly does. I'm going to assume that whatever is happening on Octane on command line is the same as what is going on in SunSpider in-browser. 1117033 8 fpizlo 2015-08-12 13:08:25 -0700 Benchmark report for Octane on shakezilla (MacBookPro11,3). VMs tested: "TipOfTree" at /Volumes/Data/secondary/OpenSource/WebKitBuild/Release/jsc (r188348) "FixGetByOffset" at /Volumes/Data/primary/OpenSource/WebKitBuild/Release/jsc (r188348) Collected 6 samples per benchmark/VM, with 6 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 FixGetByOffset encrypt 0.20010+-0.00681 ? 0.20493+-0.00225 ? might be 1.0242x slower decrypt 3.29443+-0.01743 ? 3.31550+-0.02264 ? deltablue x2 0.15490+-0.00419 0.15388+-0.00284 earley 0.27660+-0.00404 0.27627+-0.00071 boyer 4.13365+-0.03254 4.13169+-0.02740 navier-stokes x2 4.97103+-0.04141 4.92903+-0.02771 raytrace x2 1.02157+-0.06426 ! 1.87176+-0.06099 ! definitely 1.8322x slower richards x2 0.10082+-0.00120 ? 0.10092+-0.00144 ? splay x2 0.33858+-0.00391 ? 0.34019+-0.00946 ? regexp x2 24.95480+-0.30264 24.64851+-0.31738 might be 1.0124x faster pdfjs x2 37.27119+-0.25281 ? 37.48371+-0.30576 ? mandreel x2 44.23657+-0.44643 ? 44.41746+-0.73037 ? gbemu x2 34.26305+-0.57741 ? 35.12642+-2.88043 ? might be 1.0252x slower closure 0.56191+-0.00496 0.55811+-0.00253 jquery 7.11414+-0.10498 7.06744+-0.05498 box2d x2 10.15449+-0.09647 10.01343+-0.15831 might be 1.0141x faster zlib x2 372.81171+-20.02247 ? 386.93382+-10.61960 ? might be 1.0379x slower typescript x2 667.24166+-15.58583 657.88420+-14.08641 might be 1.0142x faster <geometric> 5.59631+-0.04191 ! 5.83831+-0.04758 ! definitely 1.0432x slower 1117059 9 fpizlo 2015-08-12 14:01:57 -0700 (In reply to comment #8) > Benchmark report for Octane on shakezilla (MacBookPro11,3). > > VMs tested: > "TipOfTree" at /Volumes/Data/secondary/OpenSource/WebKitBuild/Release/jsc > (r188348) > "FixGetByOffset" at /Volumes/Data/primary/OpenSource/WebKitBuild/Release/jsc > (r188348) > > Collected 6 samples per benchmark/VM, with 6 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 FixGetByOffset > > > encrypt 0.20010+-0.00681 ? 0.20493+-0.00225 ? > might be 1.0242x slower > decrypt 3.29443+-0.01743 ? 3.31550+-0.02264 ? > deltablue x2 0.15490+-0.00419 0.15388+-0.00284 > earley 0.27660+-0.00404 0.27627+-0.00071 > boyer 4.13365+-0.03254 4.13169+-0.02740 > navier-stokes x2 4.97103+-0.04141 4.92903+-0.02771 > raytrace x2 1.02157+-0.06426 ! 1.87176+-0.06099 ! > definitely 1.8322x slower > richards x2 0.10082+-0.00120 ? 0.10092+-0.00144 ? > splay x2 0.33858+-0.00391 ? 0.34019+-0.00946 ? > regexp x2 24.95480+-0.30264 24.64851+-0.31738 > might be 1.0124x faster > pdfjs x2 37.27119+-0.25281 ? 37.48371+-0.30576 ? > mandreel x2 44.23657+-0.44643 ? 44.41746+-0.73037 ? > gbemu x2 34.26305+-0.57741 ? 35.12642+-2.88043 ? > might be 1.0252x slower > closure 0.56191+-0.00496 0.55811+-0.00253 > jquery 7.11414+-0.10498 7.06744+-0.05498 > box2d x2 10.15449+-0.09647 10.01343+-0.15831 > might be 1.0141x faster > zlib x2 372.81171+-20.02247 ? 386.93382+-10.61960 ? > might be 1.0379x slower > typescript x2 667.24166+-15.58583 657.88420+-14.08641 > might be 1.0142x faster > > <geometric> 5.59631+-0.04191 ! 5.83831+-0.04758 ! > definitely 1.0432x slower Fixed in http://trac.webkit.org/changeset/188357 258735 2015-08-11 12:19:36 -0700 2015-08-11 13:01:38 -0700 the patch blah.patch text/plain 5376 fpizlo SW5kZXg6IFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291 cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkocmV2aXNpb24gMTg4MjcwKQorKysgU291cmNl L0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDE5IEBA CisyMDE1LTA4LTExICBGaWxpcCBQaXpsbyAgPGZwaXpsb0BhcHBsZS5jb20+CisKKyAgICAgICAg REZHOjpCeXRlQ29kZVBhcnNlciBzaG91bGRuJ3QgY2FsbCB0cnlHZXRDb25zdGFudFByb3BlcnR5 KCkgd2l0aCBzb21lIFN0cnVjdHVyZVNldCBpZiBpdCBpc24ndCBjaGVja2luZyB0aGF0IHRoZSBi YXNlIGhhcyBhIHN0cnVjdHVyZSBpbiB0aGF0IFN0cnVjdHVyZVNldAorICAgICAgICBodHRwczov L2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MTQ3ODkxCisgICAgICAgIHJkYXI6Ly9w cm9ibGVtLzIyMTI5NDQ3CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisK KyAgICAgICAgKiBkZmcvREZHQnl0ZUNvZGVQYXJzZXIuY3BwOgorICAgICAgICAoSlNDOjpERkc6 OkJ5dGVDb2RlUGFyc2VyOjpoYW5kbGVHZXRCeU9mZnNldCk6IEdldCByaWQgb2YgdGhpcy4KKyAg ICAgICAgKEpTQzo6REZHOjpCeXRlQ29kZVBhcnNlcjo6bG9hZCk6IERvbid0IGNhbGwgdGhlIHZl cnNpb24gb2YgaGFuZGxlR2V0QnlPZmZzZXQoKSB0aGF0IGFzc3VtZXMgdGhhdCB3ZSBoYWQgQ2hl Y2tTdHJ1Y3R1cmUnZCBzb21lIFN0cnVjdHVyZVNldCwgc2luY2Ugd2UgbWF5IG5vdCBoYXZlIENo ZWNrU3RydWN0dXJlJ2QgYW55dGhpbmcuCisgICAgICAgICogZGZnL0RGR0dyYXBoLmNwcDoKKyAg ICAgICAgKEpTQzo6REZHOjpHcmFwaDo6YXNzZXJ0SXNSZWdpc3RlcmVkKTogTWFrZSB0aGlzIGFs d2F5cyBhc3NlcnQgZXZlbiBiZWZvcmUgdGhlIFN0cnVjdHVyZVJlZ2lzdHJhdGlvblBoYXNlLgor ICAgICAgICAqIGRmZy9ERkdTdHJ1Y3R1cmVSZWdpc3RyYXRpb25QaGFzZS5jcHA6CisgICAgICAg IChKU0M6OkRGRzo6U3RydWN0dXJlUmVnaXN0cmF0aW9uUGhhc2U6OnJ1bik6IEFkZCBhIEZJWE1F IHRoYXQgbm90ZXMgdGhhdCB3ZSBubyBsb25nZXIgYmVsaWV2ZSB0aGF0IHN0cnVjdHVyZXMgc2hv dWxkIGJlIHJlZ2lzdGVyZWQgb25seSBhdCB0aGlzIHBoYXNlLiBUaGV5IHNob3VsZCBiZSByZWdp c3RlcmVkIGJlZm9yZSB0aGlzIHBoYXNlIGFuZCB0aGlzIHBoYXNlIHNob3VsZCBiZSByZW1vdmVk LgorCiAyMDE1LTA4LTExICBZdXN1a2UgU3V6dWtpICA8dXRhdGFuZS50ZWFAZ21haWwuY29tPgog CiAgICAgICAgIE51bWVyaWMgc2V0dGVyIG9uIHByb3RvdHlwZSBkb2Vzbid0IGdldCBjYWxsZWQu CkluZGV4OiBTb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR0J5dGVDb2RlUGFyc2VyLmNwcAo9 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09Ci0tLSBTb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR0J5dGVDb2RlUGFyc2Vy LmNwcAkocmV2aXNpb24gMTg4MjY2KQorKysgU291cmNlL0phdmFTY3JpcHRDb3JlL2RmZy9ERkdC eXRlQ29kZVBhcnNlci5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTIwNCw3ICsyMDQsNiBAQCBwcml2 YXRlOgogICAgIGJvb2wgaGFuZGxlQ29uc3RhbnRJbnRlcm5hbEZ1bmN0aW9uKGludCByZXN1bHRP cGVyYW5kLCBJbnRlcm5hbEZ1bmN0aW9uKiwgaW50IHJlZ2lzdGVyT2Zmc2V0LCBpbnQgYXJndW1l bnRDb3VudEluY2x1ZGluZ1RoaXMsIENvZGVTcGVjaWFsaXphdGlvbktpbmQsIGNvbnN0IENoZWNr c0Z1bmN0b3ImIGluc2VydENoZWNrcyk7CiAgICAgTm9kZSogaGFuZGxlUHV0QnlPZmZzZXQoTm9k ZSogYmFzZSwgdW5zaWduZWQgaWRlbnRpZmllciwgUHJvcGVydHlPZmZzZXQsIE5vZGUqIHZhbHVl KTsKICAgICBOb2RlKiBoYW5kbGVHZXRCeU9mZnNldChTcGVjdWxhdGVkVHlwZSwgTm9kZSogYmFz ZSwgdW5zaWduZWQgaWRlbnRpZmllck51bWJlciwgUHJvcGVydHlPZmZzZXQsIE5vZGVUeXBlID0g R2V0QnlPZmZzZXQpOwotICAgIE5vZGUqIGhhbmRsZUdldEJ5T2Zmc2V0KFNwZWN1bGF0ZWRUeXBl LCBOb2RlKiBiYXNlLCBjb25zdCBTdHJ1Y3R1cmVTZXQmLCB1bnNpZ25lZCBpZGVudGlmaWVyTnVt YmVyLCBQcm9wZXJ0eU9mZnNldCwgTm9kZVR5cGUgPSBHZXRCeU9mZnNldCk7CiAgICAgTm9kZSog aGFuZGxlR2V0QnlPZmZzZXQoU3BlY3VsYXRlZFR5cGUsIE5vZGUqIGJhc2UsIFVuaXF1ZWRTdHJp bmdJbXBsKiwgUHJvcGVydHlPZmZzZXQsIE5vZGVUeXBlID0gR2V0QnlPZmZzZXQpOwogCiAgICAg Ly8gQ3JlYXRlIGEgcHJlc2VuY2UgT2JqZWN0UHJvcGVydHlDb25kaXRpb24gYmFzZWQgb24gc29t ZSBrbm93biBvZmZzZXQgYW5kIHN0cnVjdHVyZSBzZXQuIERvZXMgbm90CkBAIC0yMjg2LDE4ICsy Mjg1LDYgQEAgTm9kZSogQnl0ZUNvZGVQYXJzZXI6OmhhbmRsZUdldEJ5T2Zmc2V0KAogICAgIHJl dHVybiBnZXRCeU9mZnNldDsKIH0KIAotTm9kZSogQnl0ZUNvZGVQYXJzZXI6OmhhbmRsZUdldEJ5 T2Zmc2V0KFNwZWN1bGF0ZWRUeXBlIHByZWRpY3Rpb24sIE5vZGUqIGJhc2UsIGNvbnN0IFN0cnVj dHVyZVNldCYgc3RydWN0dXJlU2V0LCB1bnNpZ25lZCBpZGVudGlmaWVyTnVtYmVyLCBQcm9wZXJ0 eU9mZnNldCBvZmZzZXQsIE5vZGVUeXBlIG9wKQotewotICAgIGlmIChiYXNlLT5oYXNDb25zdGFu dCgpKSB7Ci0gICAgICAgIGlmIChKU1ZhbHVlIGNvbnN0YW50ID0gbV9ncmFwaC50cnlHZXRDb25z dGFudFByb3BlcnR5KGJhc2UtPmFzSlNWYWx1ZSgpLCBzdHJ1Y3R1cmVTZXQsIG9mZnNldCkpIHsK LSAgICAgICAgICAgIGFkZFRvR3JhcGgoUGhhbnRvbSwgYmFzZSk7Ci0gICAgICAgICAgICByZXR1 cm4gd2Vha0pTQ29uc3RhbnQoY29uc3RhbnQpOwotICAgICAgICB9Ci0gICAgfQotICAgIAotICAg IHJldHVybiBoYW5kbGVHZXRCeU9mZnNldChwcmVkaWN0aW9uLCBiYXNlLCBpZGVudGlmaWVyTnVt YmVyLCBvZmZzZXQsIG9wKTsKLX0KLQogTm9kZSogQnl0ZUNvZGVQYXJzZXI6OmhhbmRsZVB1dEJ5 T2Zmc2V0KE5vZGUqIGJhc2UsIHVuc2lnbmVkIGlkZW50aWZpZXIsIFByb3BlcnR5T2Zmc2V0IG9m ZnNldCwgTm9kZSogdmFsdWUpCiB7CiAgICAgTm9kZSogcHJvcGVydHlTdG9yYWdlOwpAQCAtMjU5 Myw4ICsyNTgwLDcgQEAgTm9kZSogQnl0ZUNvZGVQYXJzZXI6OmxvYWQoCiAgICAgICAgIGxvYWRl ZFZhbHVlID0gbG9hZChsb2FkUHJlZGljdGlvbiwgdmFyaWFudC5jb25kaXRpb25TZXQoKSwgbG9h ZE9wKTsKICAgICBlbHNlIHsKICAgICAgICAgbG9hZGVkVmFsdWUgPSBoYW5kbGVHZXRCeU9mZnNl dCgKLSAgICAgICAgICAgIGxvYWRQcmVkaWN0aW9uLCBiYXNlLCB2YXJpYW50LnN0cnVjdHVyZVNl dCgpLCBpZGVudGlmaWVyTnVtYmVyLCB2YXJpYW50Lm9mZnNldCgpLAotICAgICAgICAgICAgbG9h ZE9wKTsKKyAgICAgICAgICAgIGxvYWRQcmVkaWN0aW9uLCBiYXNlLCBpZGVudGlmaWVyTnVtYmVy LCB2YXJpYW50Lm9mZnNldCgpLCBsb2FkT3ApOwogICAgIH0KIAogICAgIHJldHVybiBsb2FkZWRW YWx1ZTsKSW5kZXg6IFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9kZmcvREZHR3JhcGguY3BwCj09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT0KLS0tIFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9kZmcvREZHR3JhcGguY3BwCShyZXZpc2lv biAxODgyNjYpCisrKyBTb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR0dyYXBoLmNwcAkod29y a2luZyBjb3B5KQpAQCAtMTI4MCw5ICsxMjgwLDYgQEAgdm9pZCBHcmFwaDo6YXNzZXJ0SXNSZWdp c3RlcmVkKFN0cnVjdHVyZQogICAgIGlmICghc3RydWN0dXJlKQogICAgICAgICByZXR1cm47CiAg ICAgCi0gICAgaWYgKG1fc3RydWN0dXJlUmVnaXN0cmF0aW9uU3RhdGUgPT0gSGF2ZU5vdFN0YXJ0 ZWRSZWdpc3RlcmluZykKLSAgICAgICAgcmV0dXJuOwotICAgIAogICAgIERGR19BU1NFUlQoKnRo aXMsIG51bGxwdHIsIG1fcGxhbi53ZWFrUmVmZXJlbmNlcy5jb250YWlucyhzdHJ1Y3R1cmUpKTsK ICAgICAKICAgICBpZiAoIXN0cnVjdHVyZS0+ZGZnU2hvdWxkV2F0Y2goKSkKSW5kZXg6IFNvdXJj ZS9KYXZhU2NyaXB0Q29yZS9kZmcvREZHU3RydWN0dXJlUmVnaXN0cmF0aW9uUGhhc2UuY3BwCj09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT0KLS0tIFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9kZmcvREZHU3RydWN0dXJlUmVnaXN0 cmF0aW9uUGhhc2UuY3BwCShyZXZpc2lvbiAxODgyNjYpCisrKyBTb3VyY2UvSmF2YVNjcmlwdENv cmUvZGZnL0RGR1N0cnVjdHVyZVJlZ2lzdHJhdGlvblBoYXNlLmNwcAkod29ya2luZyBjb3B5KQpA QCAtNDQsNiArNDQsMTMgQEAgcHVibGljOgogICAgIAogICAgIGJvb2wgcnVuKCkKICAgICB7Cisg ICAgICAgIC8vIEZJWE1FOiBUaGlzIHBoYXNlIHNob3VsZG4ndCBleGlzdC4gV2Ugc2hvdWxkIGhh dmUgcmVnaXN0ZXJlZCBhbGwgc3RydWN0dXJlcyBieSBub3csIHNpbmNlCisgICAgICAgIC8vIHdl IG1heSBhbHJlYWR5IGhhdmUgZG9uZSBvcHRpbWl6YXRpb25zIHRoYXQgcmVseSBvbiBzdHJ1Y3R1 cmVzIGhhdmluZyBiZWVuIHJlZ2lzdGVyZWQuCisgICAgICAgIC8vIEN1cnJlbnRseSwgd2Ugc3Rp bGwgaGF2ZSBwbGFjZXMgd2hlcmUgd2UgZG9uJ3QgcmVnaXN0ZXIgc3RydWN0dXJlcyBwcmlvciB0 byB0aGlzIHBoYXNlLAorICAgICAgICAvLyBidXQgc3RydWN0dXJlcyBkb24ndCBlbmQgdXAgYmVp bmcgdXNlZCBmb3Igb3B0aW1pemF0aW9uIHByaW9yIHRvIHRoaXMgcGhhc2UuIFRoYXQncyBhCisg ICAgICAgIC8vIHByZXR0eSBmcmFnaWxlIHNpdHVhdGlvbiBhbmQgd2Ugc2hvdWxkIGZpeCBpdCBl dmVudHVhbGx5LgorICAgICAgICAvLyBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5j Z2k/aWQ9MTQ3ODg5CisgICAgICAgIAogICAgICAgICAvLyBXZSBuZWVkIHRvIHNldCB0aGlzIGJl Zm9yZSB0aGlzIHBoYXNlIGZpbmlzaGVzLiBUaGlzIHBoYXNlIGRvZXNuJ3QgZG8gYW55dGhpbmcK ICAgICAgICAgLy8gY29uZGl0aW9uZWQgb24gdGhpcyBmaWVsZCwgZXhjZXB0IGZvciBhc3NlcnRJ c1JlZ2lzdGVyZWQoKSBiZWxvdy4gV2UgaW50ZW5kIGZvciB0aGF0CiAgICAgICAgIC8vIG1ldGhv ZCB0byBiZWhhdmUgYXMgaWYgdGhlIHBoYXNlIHdhcyBhbHJlYWR5IGZpbmlzaGVkLiBTbywgd2Ug c2V0IHRoaXMgdXAgaGVyZS4K