111902 2013-03-08 16:51:42 -0800 ASSERT d->m_defersLoading != defers on detik.com and drive.google.com 2013-10-23 15:57:14 -0700 1 1 1 Unclassified WebKit Page Loading 528+ (Nightly build) Unspecified Unspecified RESOLVED FIXED InRadar P2 Normal --- 1 ap japhet abarth beidson japhet jochen koivisto psolanki simon.fraser webkit.review.bot oldest_to_newest 851619 0 ap 2013-03-08 16:51:42 -0800 I can fairly reliably reproduce this assertion failing on detik.com with just a few reloads. I didn't investigate in a lot of detail, but looks like the page has multiple iframes with the same URL. Their MainResourceLoaders all appear to share the same ResourceHandle. As every iframe begins to load, Safari briefly defers loads for the MainResourceLoader. If another load starts and gets deferred in the meanwhile, then we are deferring the same ResourceHandle again. 851621 1 ap 2013-03-08 16:51:52 -0800 <rdar://problem/13267540> 856335 2 ap 2013-03-15 15:21:57 -0700 I'm also constantly hitting this on drive.google.com. Nate, this sounds like a regression from caching main resources. 856347 3 japhet 2013-03-15 15:26:14 -0700 (In reply to comment #2) > I'm also constantly hitting this on drive.google.com. > > Nate, this sounds like a regression from caching main resources. (In reply to comment #2) > I'm also constantly hitting this on drive.google.com. > > Nate, this sounds like a regression from caching main resources. Hmm....that's odd. I suppose we could always just check to see if we're already deferred in MainResourceLoader::setDefersLoading() and ignore if we're trying to set to the same value. I'm not sure if that's too hackish though. 856374 4 ap 2013-03-15 16:05:37 -0700 What make this difficult for me is that I don't understand all use cases for deferring. A safe browsing check doesn't seem to require deferring an ongoing load - it can happen either from willSendRequest, or before transitioning to committed state. Do we ever actually need to defer an ongoing load? 856665 5 ap 2013-03-16 22:15:00 -0700 Or perhaps we should just count deferrals. I don't know why they aren't counted. 856741 6 simon.fraser 2013-03-17 13:32:39 -0700 fast/frames/sandboxed-iframe-navigation-allowed.html hits this sometimes: http://build.webkit.org/results/Apple%20Lion%20Debug%20WK1%20(Tests)/r146019%20(7503)/fast/frames/sandboxed-iframe-navigation-allowed-crash-log.txt ASSERTION FAILED: d->m_defersLoading != defers /Volumes/Data/slave/lion-debug/build/Source/WebCore/platform/network/ResourceHandle.cpp(202) : void WebCore::ResourceHandle::setDefersLoading(bool) 1 0x10bafe9c5 WebCore::ResourceHandle::setDefersLoading(bool) 2 0x10bb0ac21 WebCore::ResourceLoader::setDefersLoading(bool) 3 0x10b65a6c6 WebCore::MainResourceLoader::setDefersLoading(bool) 4 0x10a8fe80d WebCore::DocumentLoader::setDefersLoading(bool) 5 0x10abd96fd WebCore::FrameLoader::setDefersLoading(bool) 6 0x10b74b9ea WebCore::Page::setDefersLoading(bool) 7 0x109dfcdd0 -[WebBaseNetscapePluginView start] 8 0x109dfd726 -[WebBaseNetscapePluginView viewDidMoveToWindow] 9 0x7fff878923a5 -[NSView _setWindow:] 10 0x7fff878905bd -[NSView addSubview:] 11 0x109e86a79 -[WebHTMLView addSubview:] 12 0x10bbb36b4 WebCore::ScrollView::platformAddChild(WebCore::Widget*) 13 0x10bba9f23 WebCore::ScrollView::addChild(WTF::PassRefPtr<WebCore::Widget>) 14 0x10bae9bd6 _ZN7WebCoreL22moveWidgetToParentSoonEPNS_6WidgetEPNS_9FrameViewE 15 0x10bae9b58 WebCore::RenderWidget::setWidget(WTF::PassRefPtr<WebCore::Widget>) 16 0x10b9c1893 WebCore::RenderPart::setWidget(WTF::PassRefPtr<WebCore::Widget>) 17 0x10bd26127 WebCore::SubframeLoader::loadPlugin(WebCore::HTMLPlugInImageElement*, WebCore::KURL const&, WTF::String const&, WTF::Vector<WTF::String, 0ul> const&, WTF::Vector<WTF::String, 0ul> const&, bool) 18 0x10bd25d79 WebCore::SubframeLoader::requestPlugin(WebCore::HTMLPlugInImageElement*, WebCore::KURL const&, WTF::String const&, WTF::Vector<WTF::String, 0ul> const&, WTF::Vector<WTF::String, 0ul> const&, bool) 19 0x10bd26349 WebCore::SubframeLoader::requestObject(WebCore::HTMLPlugInImageElement*, WTF::String const&, WTF::AtomicString const&, WTF::String const&, WTF::Vector<WTF::String, 0ul> const&, WTF::Vector<WTF::String, 0ul> const&) 20 0x10adebc98 WebCore::HTMLObjectElement::updateWidget(WebCore::PluginCreationOption) 21 0x10ac0e90c WebCore::FrameView::updateWidget(WebCore::RenderObject*) 22 0x10ac0eb59 WebCore::FrameView::updateWidgets() 23 0x10ac08f4d WebCore::FrameView::performPostLayoutTasks() 24 0x10ac086f9 WebCore::FrameView::layout(bool) 25 0x10a89b2a2 WebCore::Document::implicitClose() 26 0x10abdcdab WebCore::FrameLoader::checkCallImplicitClose() 27 0x10abdca56 WebCore::FrameLoader::checkCompleted() 28 0x10abdb6b3 WebCore::FrameLoader::finishedParsing() 29 0x10a8a5f79 WebCore::Document::finishedParsing() 30 0x10ad47dec WebCore::HTMLConstructionSite::finishedParsing() 31 0x10ae3bceb WebCore::HTMLTreeBuilder::finished() 857464 7 simon.fraser 2013-03-18 13:22:25 -0700 Also fast/frames/sandboxed-iframe-plugins.html http://build.webkit.org/results/Apple%20Lion%20Debug%20WK1%20(Tests)/r146095%20(7528)/results.html http://build.webkit.org/results/Apple%20Lion%20Debug%20WK1%20(Tests)/r146095%20(7528)/fast/frames/sandboxed-iframe-plugins-crash-log.txt 860020 8 simon.fraser 2013-03-20 22:03:23 -0700 Easy to hit loading techradar.com/us 860372 9 japhet 2013-03-21 08:36:42 -0700 (In reply to comment #4) > What make this difficult for me is that I don't understand all use cases for deferring. A safe browsing check doesn't seem to require deferring an ongoing load - it can happen either from willSendRequest, or before transitioning to committed state. > > Do we ever actually need to defer an ongoing load? Does any platform's ResourceHandle support deferring a load already in progress? I'm pretty confident that WebCore/loader assumes that deferral can't happen after the request is sent. 860394 10 ap 2013-03-21 08:55:28 -0700 ResourceHandleMac supports that. I don't know what we actually use, but generally speaking, the rest of WebCore shouldn't care - deferring callbacks is indistinguishable from just having a slow connection. 860399 11 japhet 2013-03-21 09:03:55 -0700 (In reply to comment #10) > ResourceHandleMac supports that. I don't know what we actually use, but generally speaking, the rest of WebCore shouldn't care - deferring callbacks is indistinguishable from just having a slow connection. Fair point. I'm happy to work on this, I'm just not clear what the preferred solution is. Should we put in a stopgap to prevent re-deferring? 860402 12 jochen 2013-03-21 09:10:54 -0700 (In reply to comment #11) > (In reply to comment #10) > > ResourceHandleMac supports that. I don't know what we actually use, but generally speaking, the rest of WebCore shouldn't care - deferring callbacks is indistinguishable from just having a slow connection. > > Fair point. > > I'm happy to work on this, I'm just not clear what the preferred solution is. Should we put in a stopgap to prevent re-deferring? Nate, that's the assert I was talking to you about on Monday (which I hit a lot when running layout tests in content shell) 864007 13 195164 japhet 2013-03-26 14:30:52 -0700 Created attachment 195164 patch 866231 14 195164 ap 2013-03-29 10:52:36 -0700 Comment on attachment 195164 patch I guess we can try solving it this way. Unsure if anyone understands what this "deferring" is all about, but if they do, they are keeping the knowledge to themselves. 866237 15 195164 japhet 2013-03-29 10:59:01 -0700 Comment on attachment 195164 patch Yeah... :/ 866245 16 195164 webkit.review.bot 2013-03-29 11:09:00 -0700 Comment on attachment 195164 patch Clearing flags on attachment: 195164 Committed r147228: <http://trac.webkit.org/changeset/147228> 866246 17 webkit.review.bot 2013-03-29 11:09:05 -0700 All reviewed patches have been landed. Closing bug. 942877 18 ap 2013-10-23 15:57:14 -0700 Updated TestExpectations in r157896. 195164 2013-03-26 14:30:52 -0700 2013-03-29 11:08:59 -0700 patch sdl.txt text/plain 4010 japhet SW5kZXg6IExheW91dFRlc3RzL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBMYXlvdXRUZXN0cy9D aGFuZ2VMb2cJKHJldmlzaW9uIDE0NjkyNikKKysrIExheW91dFRlc3RzL0NoYW5nZUxvZwkod29y a2luZyBjb3B5KQpAQCAtMSwzICsxLDEzIEBACisyMDEzLTAzLTI2ICBOYXRlIENoYXBpbiAgPGph cGhldEBjaHJvbWl1bS5vcmc+CisKKyAgICAgICAgQVNTRVJUIGQtPm1fZGVmZXJzTG9hZGluZyAh PSBkZWZlcnMgb24gZGV0aWsuY29tIGFuZCBkcml2ZS5nb29nbGUuY29tCisgICAgICAgIGh0dHBz Oi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xMTE5MDIKKworICAgICAgICBSZXZp ZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICAqIGh0dHAvdGVzdHMvbmF2aWdhdGlv bi9zYW1lLXVybC1pZnJhbWVzLWRlZmVyLWNyYXNoLWV4cGVjdGVkLnR4dDogQWRkZWQuCisgICAg ICAgICogaHR0cC90ZXN0cy9uYXZpZ2F0aW9uL3NhbWUtdXJsLWlmcmFtZXMtZGVmZXItY3Jhc2gu aHRtbDogQWRkZWQuCisKIDIwMTMtMDMtMjYgIEFsb2sgUHJpeWFkYXJzaGkgIDxhbG9rcEBjaHJv bWl1bS5vcmc+CiAKICAgICAgICAgUmViYXNlbGluZSB0ZXh0IG91dHB1dCBmb3Igb3BhcXVlIEdy YXBoaWNzTGF5ZXJzCkluZGV4OiBMYXlvdXRUZXN0cy9odHRwL3Rlc3RzL25hdmlnYXRpb24vc2Ft ZS11cmwtaWZyYW1lcy1kZWZlci1jcmFzaC1leHBlY3RlZC50eHQKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5 b3V0VGVzdHMvaHR0cC90ZXN0cy9uYXZpZ2F0aW9uL3NhbWUtdXJsLWlmcmFtZXMtZGVmZXItY3Jh c2gtZXhwZWN0ZWQudHh0CShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMvaHR0cC90ZXN0cy9u YXZpZ2F0aW9uL3NhbWUtdXJsLWlmcmFtZXMtZGVmZXItY3Jhc2gtZXhwZWN0ZWQudHh0CSh3b3Jr aW5nIGNvcHkpCkBAIC0wLDAgKzEsMyBAQAorQUxFUlQ6IFBBU1MKK1RoaXMgdGVzdHMgdGhhdCB3 ZSBjYW4gY2F1c2UgdGhlIFBhZ2UgdG8gZGVmZXIgbG9hZGluZyB3aGlsZSBsb2FkaW5nIHRoZSBz YW1lIHJlc291cmNlIGluIG11bHRpcGxlIGlmcmFtZXMuIEluIHRoaXMgdGVzdCwgdGhlIGxvYWQg ZGVmZXJyYWwgaXMgYmVjYXVzZSBvZiBhIG1vZGFsIGRpYWxvZyB2aWEgd2luZG93LmFsZXJ0LiBX ZSBwYXNzIGlmIHdlIGRvbid0IGFzc2VydCBpbiBkZWJ1Zy4KK1NlZSBodHRwczovL2J1Z3Mud2Vi a2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MTExOTAyLiAgCkluZGV4OiBMYXlvdXRUZXN0cy9odHRw L3Rlc3RzL25hdmlnYXRpb24vc2FtZS11cmwtaWZyYW1lcy1kZWZlci1jcmFzaC5odG1sCj09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT0KLS0tIExheW91dFRlc3RzL2h0dHAvdGVzdHMvbmF2aWdhdGlvbi9zYW1lLXVybC1pZnJh bWVzLWRlZmVyLWNyYXNoLmh0bWwJKHJldmlzaW9uIDApCisrKyBMYXlvdXRUZXN0cy9odHRwL3Rl c3RzL25hdmlnYXRpb24vc2FtZS11cmwtaWZyYW1lcy1kZWZlci1jcmFzaC5odG1sCSh3b3JraW5n IGNvcHkpCkBAIC0wLDAgKzEsMTEgQEAKKzxib2R5PgorVGhpcyB0ZXN0cyB0aGF0IHdlIGNhbiBj YXVzZSB0aGUgUGFnZSB0byBkZWZlciBsb2FkaW5nIHdoaWxlIGxvYWRpbmcgdGhlIHNhbWUgcmVz b3VyY2UgaW4gbXVsdGlwbGUgaWZyYW1lcy4gSW4gdGhpcyB0ZXN0LCB0aGUgbG9hZCBkZWZlcnJh bCBpcyBiZWNhdXNlIG9mIGEgbW9kYWwgZGlhbG9nIHZpYSB3aW5kb3cuYWxlcnQuIFdlIHBhc3Mg aWYgd2UgZG9uJ3QgYXNzZXJ0IGluIGRlYnVnLjxicj4KK1NlZSBodHRwczovL2J1Z3Mud2Via2l0 Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MTExOTAyLgorPGlmcmFtZSBzcmM9InJlc291cmNlcy9ibGFu ay50eHQiPjwvaWZyYW1lPgorPGlmcmFtZSBzcmM9InJlc291cmNlcy9ibGFuay50eHQiPjwvaWZy YW1lPgorPHNjcmlwdD4KK2lmICh3aW5kb3cudGVzdFJ1bm5lcikKKyAgICB0ZXN0UnVubmVyLmR1 bXBBc1RleHQoKTsKK2FsZXJ0KCJQQVNTIik7Cis8L3NjcmlwdD4KKzwvYm9keT4KSW5kZXg6IFNv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2ViQ29yZS9DaGFu Z2VMb2cJKHJldmlzaW9uIDE0NjkyNikKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwkod29y a2luZyBjb3B5KQpAQCAtMSwzICsxLDE3IEBACisyMDEzLTAzLTI2ICBOYXRlIENoYXBpbiAgPGph cGhldEBjaHJvbWl1bS5vcmc+CisKKyAgICAgICAgQVNTRVJUIGQtPm1fZGVmZXJzTG9hZGluZyAh PSBkZWZlcnMgb24gZGV0aWsuY29tIGFuZCBkcml2ZS5nb29nbGUuY29tCisgICAgICAgIGh0dHBz Oi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xMTE5MDIKKworICAgICAgICBSZXZp ZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBUZXN0OiBodHRwL3Rlc3RzL25hdmln YXRpb24vc2FtZS11cmwtaWZyYW1lcy1kZWZlci1jcmFzaC5odG1sCisKKyAgICAgICAgKiBsb2Fk ZXIvRG9jdW1lbnRMb2FkZXIuY3BwOgorICAgICAgICAoV2ViQ29yZTo6RG9jdW1lbnRMb2FkZXI6 OnNldERlZmVyc0xvYWRpbmcpOiBJZiBtdWx0aXBsZSBEb2N1bWVudExvYWRlcnMgYXJlCisgICAg ICAgICAgIHVzaW5nIGxvYWRpbmcgdGhlIHNhbWUgbWFpbiByZXNvdXJjZSwgZW5zdXJlIG9ubHkg b25lIG9mIHRoZW0gY2FuIGNhbGwKKyAgICAgICAgICAgUmVzb3VyY2VMb2FkZXI6OnNldERlZmVy c0xvYWRpbmcuCisKIDIwMTMtMDMtMjYgIFRvbnkgQ2hhbmcgIDx0b255QGNocm9taXVtLm9yZz4K IAogICAgICAgICBBdXRvZ2VuZXJhdGUgdGhlIHNjcm9sbEFuaW1hdG9yRW5hYmxlZCBzZXR0aW5n IGluIFNldHRpbmdzLmluCkluZGV4OiBTb3VyY2UvV2ViQ29yZS9sb2FkZXIvRG9jdW1lbnRMb2Fk ZXIuY3BwCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT0KLS0tIFNvdXJjZS9XZWJDb3JlL2xvYWRlci9Eb2N1bWVudExvYWRl ci5jcHAJKHJldmlzaW9uIDE0NjY0MCkKKysrIFNvdXJjZS9XZWJDb3JlL2xvYWRlci9Eb2N1bWVu dExvYWRlci5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTEyNzMsOCArMTI3MywxMiBAQAogCiB2b2lk IERvY3VtZW50TG9hZGVyOjpzZXREZWZlcnNMb2FkaW5nKGJvb2wgZGVmZXJzKQogewotICAgIGlm IChtYWluUmVzb3VyY2VMb2FkZXIoKSkKKyAgICAvLyBNdWx0aXBsZSBmcmFtZXMgbWF5IGJlIGxv YWRpbmcgdGhlIHNhbWUgbWFpbiByZXNvdXJjZSBzaW11bHRhbmVvdXNseS4gSWYgZGVmZXJyYWwg c3RhdGUgY2hhbmdlcywKKyAgICAvLyBlYWNoIGZyYW1lJ3MgRG9jdW1lbnRMb2FkZXIgd2lsbCB0 cnkgdG8gc2VuZCBhIHNldERlZmVyc0xvYWRpbmcoKSB0byB0aGUgc2FtZSB1bmRlcmx5aW5nIFJl c291cmNlTG9hZGVyLiBFbnN1cmUgb25seQorICAgIC8vIHRoZSAib3duaW5nIiBEb2N1bWVudExv YWRlciBkb2VzIHNvLCBhcyBzZXREZWZlcnNMb2FkaW5nKCkgaXMgbm90IHJlc2lsaWVudCB0byBz ZXR0aW5nIHRoZSBzYW1lIHZhbHVlIHJlcGVhdGVkbHkuCisgICAgaWYgKG1haW5SZXNvdXJjZUxv YWRlcigpICYmIG1haW5SZXNvdXJjZUxvYWRlcigpLT5kb2N1bWVudExvYWRlcigpID09IHRoaXMp CiAgICAgICAgIG1haW5SZXNvdXJjZUxvYWRlcigpLT5zZXREZWZlcnNMb2FkaW5nKGRlZmVycyk7 CisKICAgICBzZXRBbGxEZWZlcnNMb2FkaW5nKG1fc3VicmVzb3VyY2VMb2FkZXJzLCBkZWZlcnMp OwogICAgIHNldEFsbERlZmVyc0xvYWRpbmcobV9wbHVnSW5TdHJlYW1Mb2FkZXJzLCBkZWZlcnMp OwogICAgIGlmICghZGVmZXJzKQo=