138440 2014-11-05 14:44:30 -0800 Assertion hit DOMTimer::updateTimerIntervalIfNecessary() 2014-11-07 15:51:36 -0800 1 1 1 Unclassified WebKit DOM 528+ (Nightly build) Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 cdumez cdumez barraclough benjamin cmarcelo commit-queue darin ggaren oldest_to_newest 1046606 0 cdumez 2014-11-05 14:44:30 -0800 I sometimes hit the following assertion in DOMTimer::updateTimerIntervalIfNecessary(): ASSERT(repeatInterval() == previousInterval); when visiting the following URLs: http://lifehacker.com/the-healthiest-foods-for-one-handed-snacking-while-gami-1654728164 http://longform.org/posts/like-something-the-lord-made Trace: Thread 0 Crashed:: Dispatch queue: com.apple.main-thread 0 com.apple.JavaScriptCore 0x000000011324634a WTFCrash + 42 1 com.apple.WebCore 0x0000000114e93f37 WebCore::DOMTimer::updateTimerIntervalIfNecessary() + 247 (DOMTimer.cpp:324) 2 com.apple.WebCore 0x0000000114e93e2c WebCore::DOMTimer::updateThrottlingStateIfNecessary(WebCore::DOMTimerFireState const&) + 140 (DOMTimer.cpp:203) 3 com.apple.WebCore 0x0000000114e942b2 WebCore::DOMTimer::fired() + 706 (DOMTimer.cpp:256) 4 com.apple.WebCore 0x00000001166b705c WebCore::ThreadTimers::sharedTimerFiredInternal() + 396 (ThreadTimers.cpp:135) 5 com.apple.WebCore 0x00000001166b6d19 WebCore::ThreadTimers::sharedTimerFired() + 25 (ThreadTimers.cpp:108) 6 com.apple.WebCore 0x000000011637aa2f WebCore::timerFired(__CFRunLoopTimer*, void*) + 31 (SharedTimerMac.mm:125) 7 com.apple.CoreFoundation 0x00007fff8eb54b44 __CFRUNLOOP_IS_CALLING_OUT_TO_A_TIMER_CALLBACK_FUNCTION__ + 20 8 com.apple.CoreFoundation 0x00007fff8eb547d3 __CFRunLoopDoTimer + 1059 9 com.apple.CoreFoundation 0x00007fff8ebc7d9d __CFRunLoopDoTimers + 301 10 com.apple.CoreFoundation 0x00007fff8eb11268 __CFRunLoopRun + 2024 11 com.apple.CoreFoundation 0x00007fff8eb10838 CFRunLoopRunSpecific + 296 12 com.apple.HIToolbox 0x00007fff8a0b03ff RunCurrentEventLoopInMode + 235 13 com.apple.HIToolbox 0x00007fff8a0b017a ReceiveNextEventCommon + 431 14 com.apple.HIToolbox 0x00007fff8a0affbb _BlockUntilNextEventMatchingListInModeWithFilter + 71 15 com.apple.AppKit 0x00007fff85f4b85d _DPSNextEvent + 964 16 com.apple.AppKit 0x00007fff85f4b010 -[NSApplication nextEventMatchingMask:untilDate:inMode:dequeue:] + 194 17 com.apple.AppKit 0x00007fff85f3ef93 -[NSApplication run] + 594 18 com.apple.AppKit 0x00007fff85f2a444 NSApplicationMain + 1832 19 libxpc.dylib 0x00007fff9167fef2 _xpc_objc_main + 793 20 libxpc.dylib 0x00007fff91681a9d xpc_main + 490 21 com.apple.WebKit.WebContent.Development 0x0000000108f5a115 main + 37 22 libdyld.dylib 0x00007fff8436d5c9 start + 1 Investigation: We are relying on floating point comparison in DOMTimer code and sometimes repeatInterval() == previousInterval returns false even though the values are roughly the same (due to precision). Printing those values shows "0.2 != 0.2". We should make sure the difference between the values is within std::numeric_limits<double>::epsilon() instead. Possible cause for the following Radar: <rdar://problem/18883681> 1046642 1 241069 cdumez 2014-11-05 15:53:53 -0800 Created attachment 241069 Patch 1046664 2 cdumez 2014-11-05 17:29:35 -0800 Removing InRadar keyword as it is a separate issue from <rdar://problem/18883681>. 1046672 3 241069 ggaren 2014-11-05 17:40:04 -0800 Comment on attachment 241069 Patch r=me 1046687 4 241069 commit-queue 2014-11-05 18:21:08 -0800 Comment on attachment 241069 Patch Clearing flags on attachment: 241069 Committed r175655: <http://trac.webkit.org/changeset/175655> 1046688 5 commit-queue 2014-11-05 18:21:15 -0800 All reviewed patches have been landed. Closing bug. 1047219 6 darin 2014-11-07 15:35:57 -0800 The "within epsilon" check only does something useful for values that are close enough to 0. It’s not generally a correct technique to use on arbitrary floating point values. It would be nice to rid our code of this technique if we can find a better solution to these problems in the future. 1047230 7 cdumez 2014-11-07 15:51:36 -0800 (In reply to comment #6) > The "within epsilon" check only does something useful for values that are > close enough to 0. It’s not generally a correct technique to use on > arbitrary floating point values. > > It would be nice to rid our code of this technique if we can find a better > solution to these problems in the future. Yes, agreed. I can update the function to use one of Knuth's formulas from: http://www.boost.org/doc/libs/1_34_0/libs/test/doc/components/test_tools/floating_point_comparison.html#Introduction I am not sure which formula would be best here but any of those would likely be better than the basic withinEpsilon() check we currently have. Any thought on this? 241069 2014-11-05 15:53:53 -0800 2014-11-05 18:21:08 -0800 Patch bug-138440-20141105155403.patch text/plain 5392 cdumez U3VidmVyc2lvbiBSZXZpc2lvbjogMTc1NjI5CmRpZmYgLS1naXQgYS9Tb3VyY2UvV1RGL0NoYW5n ZUxvZyBiL1NvdXJjZS9XVEYvQ2hhbmdlTG9nCmluZGV4IDg1Y2VlZWU5NTE5ZjY5ZTA0NGEwNmU1 YWQzNTdjOGYwZGFmZWJiNmMuLmFhMjVjZTFiY2RiYTc2Yzg4YjMzMDFhNThmODdiNzk0MDJiMmY0 NjQgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XVEYvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9XVEYvQ2hh bmdlTG9nCkBAIC0xLDMgKzEsMTUgQEAKKzIwMTQtMTEtMDUgIENocmlzIER1bWV6ICA8Y2R1bWV6 QGFwcGxlLmNvbT4KKworICAgICAgICBBc3NlcnRpb24gaGl0IERPTVRpbWVyOjp1cGRhdGVUaW1l ckludGVydmFsSWZOZWNlc3NhcnkoKQorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9z aG93X2J1Zy5jZ2k/aWQ9MTM4NDQwCisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BT ISkuCisKKyAgICAgICAgTW92ZSB0aGUgd2l0aGluRXBzaWxvbigpIGZ1bmN0aW9uIHRvIFdURiB0 byBhdm9pZCBjb2RlIGR1cGxpY2F0aW9uLgorCisgICAgICAgICogd3RmL01hdGhFeHRyYXMuaDoK KyAgICAgICAgKFdURjo6d2l0aGluRXBzaWxvbik6CisKIDIwMTQtMTEtMDMgIERlYW4gSmFja3Nv biAgPGRpbm9AYXBwbGUuY29tPgogCiAgICAgICAgIEFkZCBFTkFCTEVfRklMVEVSU19MRVZFTF8y IGZlYXR1cmUgZ3VhcmQuCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cgYi9T b3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggNzFlZDFlY2I0OTg0NDY1MjA4ZGIzMzM2NGM1 OTdjNDhlNjU0M2I0Mi4uZTQ5MDMxOTRhNWQ0YmFmNTZmMTMzYjViYTY0OWRjYzdjYzM4ZDQ5MiAx MDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9XZWJDb3Jl L0NoYW5nZUxvZwpAQCAtMSw1ICsxLDMyIEBACiAyMDE0LTExLTA1ICBDaHJpcyBEdW1leiAgPGNk dW1lekBhcHBsZS5jb20+CiAKKyAgICAgICAgQXNzZXJ0aW9uIGhpdCBET01UaW1lcjo6dXBkYXRl VGltZXJJbnRlcnZhbElmTmVjZXNzYXJ5KCkKKyAgICAgICAgaHR0cHM6Ly9idWdzLndlYmtpdC5v cmcvc2hvd19idWcuY2dpP2lkPTEzODQ0MAorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAo T09QUyEpLgorCisgICAgICAgIFdlIHNvbWV0aW1lcyBoaXQgdGhlIEFTU0VSVChyZXBlYXRJbnRl cnZhbCgpID09IHByZXZpb3VzSW50ZXJ2YWwpCisgICAgICAgIGFzc2VydGlvbiBpbiBET01UaW1l cjo6dXBkYXRlVGltZXJJbnRlcnZhbElmTmVjZXNzYXJ5KCkgd2hlbiB2aXNpdGluZworICAgICAg ICB0aGUgZm9sbG93aW5nIHBhZ2VzOgorICAgICAgICBodHRwOi8vbGlmZWhhY2tlci5jb20vdGhl LWhlYWx0aGllc3QtZm9vZHMtZm9yLW9uZS1oYW5kZWQtc25hY2tpbmctd2hpbGUtZ2FtaS0xNjU0 NzI4MTY0CisgICAgICAgIGh0dHA6Ly9sb25nZm9ybS5vcmcvcG9zdHMvbGlrZS1zb21ldGhpbmct dGhlLWxvcmQtbWFkZQorCisgICAgICAgIEFmdGVyIGRlYnVnZ2luZywgdGhlIGlzc3VlIHR1cm5l ZCBvdXQgdG8gYmUgdGhhdCB3ZSBhcmUgY29tcGFyaW5nCisgICAgICAgIGZsb2F0aW5nIHBvaW50 IG51bWJlcnMgdXNpbmcgPT0sIGFuZCB0aGUgY2hlY2sgc29tZXRpbWVzIGZhaWxzIGV2ZW4KKyAg ICAgICAgdGhvdWdoIHRoZSB2YWx1ZXMgcmVhbGx5IGNsb3NlIHRvIGVhY2ggb3RoZXIuIFRoaXMg cGF0Y2ggdXBkYXRlcyB0aGUKKyAgICAgICAgRE9NVGltZXIgY29kZSB0byB1c2UgV1RGOjp3aXRo aW5FcHNpbG9uKCkgaW5zdGVhZCBvZiBvcGVyYXRvcj09KCkKKyAgICAgICAgdG8gY29tcGFyZSB0 aGUgZmxvYXRpbmcgcG9pbnQgaW50ZXJ2YWxzLgorCisgICAgICAgIEkgY29uZmlybWVkIG1hbnVh bGx5IHRoYXQgdGhlIGFzc2VydGlvbiBpcyBubyBsb25nZXIgaGl0LgorCisgICAgICAgICogcGFn ZS9ET01UaW1lci5jcHA6CisgICAgICAgIChXZWJDb3JlOjpET01UaW1lcjo6dXBkYXRlVGltZXJJ bnRlcnZhbElmTmVjZXNzYXJ5KToKKyAgICAgICAgKiBwbGF0Zm9ybS9ncmFwaGljcy9GbG9hdFF1 YWQuY3BwOgorICAgICAgICAoV2ViQ29yZTo6RmxvYXRRdWFkOjppc1JlY3RpbGluZWFyKToKKyAg ICAgICAgKFdlYkNvcmU6OndpdGhpbkVwc2lsb24pOiBEZWxldGVkLgorCisyMDE0LTExLTA1ICBD aHJpcyBEdW1leiAgPGNkdW1lekBhcHBsZS5jb20+CisKICAgICAgICAgUmUtZXZhbHVhdGUgaWYg YSByZXBldGl0aXZlIHRpbWVyIHNob3VsZCBiZSB0aHJvdHRsZWQgdXBvbiBmaXJpbmcKICAgICAg ICAgaHR0cHM6Ly9idWdzLndlYmtpdC5vcmcvc2hvd19idWcuY2dpP2lkPTEzODMzOQogCmRpZmYg LS1naXQgYS9Tb3VyY2UvV1RGL3d0Zi9NYXRoRXh0cmFzLmggYi9Tb3VyY2UvV1RGL3d0Zi9NYXRo RXh0cmFzLmgKaW5kZXggYTQzZjVkMDYwYjE2MzRhMjc4ZGY1YTVmYTA0MjA0NjQ4M2Q4MGM1YS4u ZTMzOWMxYjE0NWU4N2FmYzY1OGQxYWJjNWJkYmEwOWIyODIxM2U4NyAxMDA2NDQKLS0tIGEvU291 cmNlL1dURi93dGYvTWF0aEV4dHJhcy5oCisrKyBiL1NvdXJjZS9XVEYvd3RmL01hdGhFeHRyYXMu aApAQCAtMzkyLDYgKzM5MiwxMiBAQCBpbmxpbmUgdW5zaWduZWQgZmFzdExvZzIodW5zaWduZWQg aSkKICAgICByZXR1cm4gbG9nMjsKIH0KIAordGVtcGxhdGUgPHR5cGVuYW1lIFQ+CitpbmxpbmUg Ym9vbCB3aXRoaW5FcHNpbG9uKFQgYSwgVCBiKQoreworICAgIHJldHVybiBzdGQ6OmFicyhhIC0g YikgPD0gc3RkOjpudW1lcmljX2xpbWl0czxUPjo6ZXBzaWxvbigpOworfQorCiB9IC8vIG5hbWVz cGFjZSBXVEYKIAogI2VuZGlmIC8vICNpZm5kZWYgV1RGX01hdGhFeHRyYXNfaApkaWZmIC0tZ2l0 IGEvU291cmNlL1dlYkNvcmUvcGFnZS9ET01UaW1lci5jcHAgYi9Tb3VyY2UvV2ViQ29yZS9wYWdl L0RPTVRpbWVyLmNwcAppbmRleCA5NGQyYjFmNDU4YWFiNTEwOTYyODdjOWQ2MzZkZTcyNDY4ZGQ3 N2ExLi5kYWQ2ODcyNjA4ZjE1MmJiMzZlMGZhYjczMTAxNTgwNjhhMjVlYjU2IDEwMDY0NAotLS0g YS9Tb3VyY2UvV2ViQ29yZS9wYWdlL0RPTVRpbWVyLmNwcAorKysgYi9Tb3VyY2UvV2ViQ29yZS9w YWdlL0RPTVRpbWVyLmNwcApAQCAtMzUsNiArMzUsNyBAQAogI2luY2x1ZGUgIlVzZXJHZXN0dXJl SW5kaWNhdG9yLmgiCiAjaW5jbHVkZSA8d3RmL0N1cnJlbnRUaW1lLmg+CiAjaW5jbHVkZSA8d3Rm L0hhc2hTZXQuaD4KKyNpbmNsdWRlIDx3dGYvTWF0aEV4dHJhcy5oPgogI2luY2x1ZGUgPHd0Zi9T dGRMaWJFeHRyYXMuaD4KIAogI2lmIFBMQVRGT1JNKElPUykKQEAgLTMxNywxMSArMzE4LDExIEBA IHZvaWQgRE9NVGltZXI6OnVwZGF0ZVRpbWVySW50ZXJ2YWxJZk5lY2Vzc2FyeSgpCiAgICAgZG91 YmxlIHByZXZpb3VzSW50ZXJ2YWwgPSBtX2N1cnJlbnRUaW1lckludGVydmFsOwogICAgIG1fY3Vy cmVudFRpbWVySW50ZXJ2YWwgPSBpbnRlcnZhbENsYW1wZWRUb01pbmltdW0oKTsKIAotICAgIGlm IChwcmV2aW91c0ludGVydmFsID09IG1fY3VycmVudFRpbWVySW50ZXJ2YWwpCisgICAgaWYgKFdU Rjo6d2l0aGluRXBzaWxvbihwcmV2aW91c0ludGVydmFsLCBtX2N1cnJlbnRUaW1lckludGVydmFs KSkKICAgICAgICAgcmV0dXJuOwogCiAgICAgaWYgKHJlcGVhdEludGVydmFsKCkpIHsKLSAgICAg ICAgQVNTRVJUKHJlcGVhdEludGVydmFsKCkgPT0gcHJldmlvdXNJbnRlcnZhbCk7CisgICAgICAg IEFTU0VSVChXVEY6OndpdGhpbkVwc2lsb24ocmVwZWF0SW50ZXJ2YWwoKSwgcHJldmlvdXNJbnRl cnZhbCkpOwogICAgICAgICBhdWdtZW50UmVwZWF0SW50ZXJ2YWwobV9jdXJyZW50VGltZXJJbnRl cnZhbCAtIHByZXZpb3VzSW50ZXJ2YWwpOwogICAgIH0gZWxzZQogICAgICAgICBhdWdtZW50Rmly ZUludGVydmFsKG1fY3VycmVudFRpbWVySW50ZXJ2YWwgLSBwcmV2aW91c0ludGVydmFsKTsKZGlm ZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL3BsYXRmb3JtL2dyYXBoaWNzL0Zsb2F0UXVhZC5jcHAg Yi9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9GbG9hdFF1YWQuY3BwCmluZGV4IGYx MTU4NmM5Nzk4NjU2MTM0NTU1ODhkZjUxMjA4MDFlNWJjMTkwMzIuLmI5NWZhYjJiZDEyZGJjMDEz NjA4NTJmNmQwNjVkYjY5ZjFlMWRhMjYgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XZWJDb3JlL3BsYXRm b3JtL2dyYXBoaWNzL0Zsb2F0UXVhZC5jcHAKKysrIGIvU291cmNlL1dlYkNvcmUvcGxhdGZvcm0v Z3JhcGhpY3MvRmxvYXRRdWFkLmNwcApAQCAtMzMsNiArMzMsNyBAQAogCiAjaW5jbHVkZSA8YWxn b3JpdGhtPgogI2luY2x1ZGUgPGxpbWl0cz4KKyNpbmNsdWRlIDx3dGYvTWF0aEV4dHJhcy5oPgog CiBuYW1lc3BhY2UgV2ViQ29yZSB7CiAKQEAgLTkwLDE1ICs5MSwxMCBAQCBGbG9hdFJlY3QgRmxv YXRRdWFkOjpib3VuZGluZ0JveCgpIGNvbnN0CiAgICAgcmV0dXJuIEZsb2F0UmVjdChsZWZ0LCB0 b3AsIHJpZ2h0IC0gbGVmdCwgYm90dG9tIC0gdG9wKTsKIH0KIAotc3RhdGljIGlubGluZSBib29s IHdpdGhpbkVwc2lsb24oZmxvYXQgYSwgZmxvYXQgYikKLXsKLSAgICByZXR1cm4gZmFicyhhIC0g YikgPCBzdGQ6Om51bWVyaWNfbGltaXRzPGZsb2F0Pjo6ZXBzaWxvbigpOwotfQotCiBib29sIEZs b2F0UXVhZDo6aXNSZWN0aWxpbmVhcigpIGNvbnN0CiB7Ci0gICAgcmV0dXJuICh3aXRoaW5FcHNp bG9uKG1fcDEueCgpLCBtX3AyLngoKSkgJiYgd2l0aGluRXBzaWxvbihtX3AyLnkoKSwgbV9wMy55 KCkpICYmIHdpdGhpbkVwc2lsb24obV9wMy54KCksIG1fcDQueCgpKSAmJiB3aXRoaW5FcHNpbG9u KG1fcDQueSgpLCBtX3AxLnkoKSkpCi0gICAgICAgIHx8ICh3aXRoaW5FcHNpbG9uKG1fcDEueSgp LCBtX3AyLnkoKSkgJiYgd2l0aGluRXBzaWxvbihtX3AyLngoKSwgbV9wMy54KCkpICYmIHdpdGhp bkVwc2lsb24obV9wMy55KCksIG1fcDQueSgpKSAmJiB3aXRoaW5FcHNpbG9uKG1fcDQueCgpLCBt X3AxLngoKSkpOworICAgIHJldHVybiAoV1RGOjp3aXRoaW5FcHNpbG9uKG1fcDEueCgpLCBtX3Ay LngoKSkgJiYgV1RGOjp3aXRoaW5FcHNpbG9uKG1fcDIueSgpLCBtX3AzLnkoKSkgJiYgV1RGOjp3 aXRoaW5FcHNpbG9uKG1fcDMueCgpLCBtX3A0LngoKSkgJiYgV1RGOjp3aXRoaW5FcHNpbG9uKG1f cDQueSgpLCBtX3AxLnkoKSkpCisgICAgICAgIHx8IChXVEY6OndpdGhpbkVwc2lsb24obV9wMS55 KCksIG1fcDIueSgpKSAmJiBXVEY6OndpdGhpbkVwc2lsb24obV9wMi54KCksIG1fcDMueCgpKSAm JiBXVEY6OndpdGhpbkVwc2lsb24obV9wMy55KCksIG1fcDQueSgpKSAmJiBXVEY6OndpdGhpbkVw c2lsb24obV9wNC54KCksIG1fcDEueCgpKSk7CiB9CiAKIGJvb2wgRmxvYXRRdWFkOjpjb250YWlu c1BvaW50KGNvbnN0IEZsb2F0UG9pbnQmIHApIGNvbnN0Cg==