103208 2012-11-25 16:34:39 -0800 Don't use ComposedShadowTreeWalker in NodeRenderingContext when SHADOW_DOM is not enabled 2014-05-14 23:12:05 -0700 1 1 1 Unclassified WebKit DOM 528+ (Nightly build) Unspecified Unspecified RESOLVED INVALID P2 Normal --- 103339 1 koivisto koivisto abarth dglazkov eric kenjibaheux kling mjs morrita ojan rniwa sam shinyak webcomponents-bugzilla webkit.review.bot oldest_to_newest 775046 0 koivisto 2012-11-25 16:34:39 -0800 It does tons of work that is not needed when SHADOW_DOM is not enabled. 775051 1 175895 koivisto 2012-11-25 16:45:59 -0800 Created attachment 175895 patch 775056 2 175895 sam 2012-11-25 16:50:24 -0800 Comment on attachment 175895 patch Egads. I'm not a fan of more #ifdefs, but a 3-4% regression from shadow DOM stuff is clearly not acceptable. You should file a follow up about the regression, since one day we may need to turn it on. 775079 3 175895 buildbot 2012-11-25 18:04:17 -0800 Comment on attachment 175895 patch Attachment 175895 did not pass mac-ews (mac): Output: http://queues.webkit.org/results/14987445 New failing tests: fast/html/details-add-summary-child-2.html fast/html/details-add-summary-1.html fast/html/details-add-details-child-1.html fast/html/details-add-summary-7.html fast/html/details-add-summary-5.html fast/html/details-add-summary-5-and-click.html fast/html/details-add-summary-6-and-click.html fast/html/details-add-child-1.html fast/html/details-add-summary-4.html fast/html/details-add-summary-6.html fast/html/details-click-controls.html fast/html/details-keyboard-show-hide.html fast/html/details-add-summary-2.html fast/html/details-add-summary-10.html fast/html/details-add-summary-10-and-click.html fast/html/details-add-summary-9.html fast/html/details-add-child-2.html fast/block/block-remove-child-delete-line-box-crash.html fast/html/details-add-details-child-2.html fast/html/details-add-summary-8-and-click.html fast/html/details-add-summary-1-and-click.html fast/html/details-add-summary-2-and-click.html fast/html/details-add-summary-7-and-click.html fast/html/details-add-summary-9-and-click.html fast/events/domactivate-sets-underlying-click-event-as-handled.html fast/html/details-add-summary-3-and-click.html fast/html/details-add-summary-4-and-click.html fast/html/details-add-summary-3.html fast/html/details-add-summary-child-1.html fast/html/details-add-summary-8.html 775355 4 koivisto 2012-11-26 04:36:43 -0800 Looks like the internal implementation of the <details> element has been made to depend on a bunch of Web Components features. This is not great. 775740 5 eric 2012-11-26 11:53:21 -0800 I suspect morrita is the <details> expert, judging from the revision log: http://trac.webkit.org/log/trunk/Source/WebCore/html/HTMLDetailsElement.cpp 776017 6 morrita 2012-11-26 16:15:17 -0800 Yes, <details> and <summary> uses Shadow DOM internally, which allowed us to kill a bunch of renderer-side code. I agree that the regression here isn't acceptable though. Can I take some time to address this? I'll look into it. The regression should be killed even if SHADOW_DOM is enabled, at least eventually. 776044 7 koivisto 2012-11-26 16:27:44 -0800 Web Components code should be fully behind a feature flag like all other experimental features. Platforms that are not interested in this feature shouldn't need to take the performance and binary size hit from it. Also it is not ok to make other parts of WebKit internals depend on experimental features. The original implementation of <details> did not have such dependency. The feature define should be renamed to WEB_COMPONENTS or similar. SHADOW_DOM is a bad name as we have tons of internal shadow DOMs that don't depend on most of this code. 776049 8 ojan 2012-11-26 16:33:15 -0800 (In reply to comment #7) > Web Components code should be fully behind a feature flag like all other experimental features. Platforms that are not interested in this feature shouldn't need to take the performance and binary size hit from it. Also it is not ok to make other parts of WebKit internals depend on experimental features. The original implementation of <details> did not have such dependency. > > The feature define should be renamed to WEB_COMPONENTS or similar. SHADOW_DOM is a bad name as we have tons of internal shadow DOMs that don't depend on most of this code. In general, I agree with this principle. In this specific case, getting rid of custom renderer code and replacing it with components seems like a big win to me. Every time we write custom renderer code for form controls, it's a source of a long tail of correctness and security bugs. Once (if) there is more general buy-in to shipping web components, I think replacing custom renderers with web components is a good direction to go. Maybe making those changes was premature though as it seems Apple doesn't necessarily want to ship web components? 776161 9 morrita 2012-11-26 18:05:24 -0800 For performance, we should definitely get rid of the regression. I'll take a look to make this patch work with <details>. For binary size, it's infeasible for now to hide whole Shadow DOM things behind a flag. It will result tons of ifdefs and hurts the code readability. Sine it replaces certain renderer implementations, it can taken as a nice-to-have WebCore plumbing. That being said, I agree that some more Shadow DOM/Web Components implementation can be moved behind the flag. That will be tracked by a bug 103339. On the flag name, please let us use SHADOW_DOM. The API standard is named like that http://dvcs.w3.org/hg/webcomponents/raw-file/tip/spec/shadow/index.html, and WEB_COMPONENTS sounds too generic anyway. There are other Web Components related APIs like custom dom elements and template . 776323 10 dglazkov 2012-11-26 21:50:54 -0800 (In reply to comment #7) > Web Components code should be fully behind a feature flag like all other experimental features. Platforms that are not interested in this feature shouldn't need to take the performance and binary size hit from it. Also it is not ok to make other parts of WebKit internals depend on experimental features. The original implementation of <details> did not have such dependency. > > The feature define should be renamed to WEB_COMPONENTS or similar. SHADOW_DOM is a bad name as we have tons of internal shadow DOMs that don't depend on most of this code. As I mentioned at this year's WebKit contributors meeting (http://webkitmemes.tumblr.com/image/21402870608), all of our internal shadow DOMs use the same plumbing. It's all been painfully specified here (http://www.w3.org/TR/shadow-dom/). SHADOW_DOM intentionally guards only the public-facing API, because most of it had been just a reshaping of the existing custom shadow DOM code we keep writing for every darn HTML element. Most of the Shadow DOM spec is just writing down lessons I learned from that refactoring. Speaking of which, the initial, hard-coded implementation of <details> was so bad that you would've cried tears of blood just looking at it. There are lots of tiny details that we figured out and made sure they worked right with insertion points, from accessibility to hit-testing, to event retargeting and rendering. I somehow doubt that there's a noticeable binary size hit. I'd bet that a proper <details> implementation in itself would've been a hefty chunk of code. I understand your frustration and I totally agree that we should fix the performance regression though. We suck and we should fix it -- and we will. And may Kling create a thousand ghastly memes featuring me and Morrita if we don't. 776596 11 koivisto 2012-11-27 04:02:43 -0800 (In reply to comment #9) > For binary size, it's infeasible for now to hide whole Shadow DOM things behind a flag. It will result tons of ifdefs and hurts the code readability. I'm not buying this. Quick testing showed that with <details> disabled it was fairly trivial and non-invasive to compile out ComposedShadowTreeWalker, HTMLContentElement, HTMLShadowElement, InsertionPoint, ContentDistributor and ContentSelectorQuery. That's a more than 1k lines of complex code. The rest of the engine seems happy with just ShadowRoot and parts of ElementShadow. I'm not sure if the cleaner looking <details> implementation is worth the complexity. I suspect it is completely opaque to many WebKit engineers. I spent long time trying to figure out what was going on to make this patch work with <details> and failed. 778744 12 mjs 2012-11-28 23:48:54 -0800 (In reply to comment #8) > (In reply to comment #7) > > Web Components code should be fully behind a feature flag like all other experimental features. Platforms that are not interested in this feature shouldn't need to take the performance and binary size hit from it. Also it is not ok to make other parts of WebKit internals depend on experimental features. The original implementation of <details> did not have such dependency. > > > > The feature define should be renamed to WEB_COMPONENTS or similar. SHADOW_DOM is a bad name as we have tons of internal shadow DOMs that don't depend on most of this code. > > In general, I agree with this principle. In this specific case, getting rid of custom renderer code and replacing it with components seems like a big win to me. Every time we write custom renderer code for form controls, it's a source of a long tail of correctness and security bugs. > > Once (if) there is more general buy-in to shipping web components, I think replacing custom renderers with web components is a good direction to go. Maybe making those changes was premature though as it seems Apple doesn't necessarily want to ship web components? I think it was premature. We're not necessarily opposed to ever shipping web components related code, but since it is still officially an experimental and optional feature, it seems wrong to me that we're effectively forced to ship large chunks of it in our binary, and to have unrelated features depend on it. I definitely see the potential, but I'd each port to be able to choose when/if to get on the bandwagon, at least until the WebKit community as a whole decides this is no longer an experimental or optional feature. 779273 13 koivisto 2012-11-29 11:59:39 -0800 I uploaded a WIP patch to bug 103208. Could people actively working on this feature please pick it up? 779279 14 koivisto 2012-11-29 12:00:17 -0800 I mean bug 103654. 800041 15 eric 2013-01-04 00:53:15 -0800 Attachment 175895 was posted by a committer and has review+, assigning to Antti Koivisto for commit. 1009460 16 kling 2014-05-14 23:12:05 -0700 Hue. 175895 2012-11-25 16:45:59 -0800 2012-11-25 18:04:17 -0800 patch shadow-dom-free-noderenderercontext-2.patch text/plain 8442 koivisto SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDEzNTY3NCkKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDMzIEBACisyMDEyLTExLTI1ICBBbnR0aSBL b2l2aXN0byAgPGFudHRpQGFwcGxlLmNvbT4KKworICAgICAgICBEb24ndCB1c2UgQ29tcG9zZWRT aGFkb3dUcmVlV2Fsa2VyIGluIE5vZGVSZW5kZXJpbmdDb250ZXh0IHdoZW4gU0hBRE9XX0RPTSBp cyBub3QgZW5hYmxlZAorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5j Z2k/aWQ9MTAzMjA4CisKKyAgICAgICAgUmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAg ICAgICAgQ29tcG9zZWRTaGFkb3dUcmVlV2Fsa2VyIGlzIHZlcnkgY29tcGxpY2F0ZWQgYW5kIHNs b3cgZXZlbiB3aGVuIGl0IGRvZXMgbm90aGluZyBtdWNoIGF0IGFsbC4gV2hlbiBTSEFET1dfRE9N IGlzIG5vdCBlbmFibGVkCisgICAgICAgIGl0IGNhbiBiZSByZWR1Y2VkIHRvIGEgZmV3IGxpbmVz IG9mIGNvZGUuCisgICAgICAgIAorICAgICAgICBUaGlzIHBhdGNoIG1vdmVzIHRoZSB1c2Ugb2Yg Q29tcG9zZWRTaGFkb3dUcmVlV2Fsa2VyIGluIE5vZGVSZW5kZXJpbmdDb250ZXh0IGJlaGluZCBm ZWF0dXJlIGRlZmluZXMgYW5kIGluY2x1ZGVzIGEgbWluaW1hbCAKKyAgICAgICAgZXF1aXZhbGVu dCBpbXBsZW1lbnRhdGlvbi4KKyAgICAgICAgCisgICAgICAgIFRoaXMgc2VlbXMgdG8gYmUgMy00 JSBwcm9ncmVzc2lvbiBpbiBEcm9tYWVvL2RvbS1tb2RpZnkvaW5uZXJIVE1MIGFuZCBoZWxwIHdp dGggYSBidW5jaCBvZiBvdGhlcnMgdG9vLgorCisgICAgICAgICogZG9tL05vZGVSZW5kZXJpbmdD b250ZXh0LmNwcDoKKyAgICAgICAgKFdlYkNvcmU6Ok5vZGVSZW5kZXJpbmdDb250ZXh0OjpOb2Rl UmVuZGVyaW5nQ29udGV4dCk6CisgICAgICAgIChXZWJDb3JlOjppc1ZhbGlkTm9uRmxvd1RocmVh ZFJlbmRlcmVyKToKKyAgICAgICAgKFdlYkNvcmUpOgorICAgICAgICAoV2ViQ29yZTo6Tm9kZVJl bmRlcmluZ0NvbnRleHQ6Om5leHRSZW5kZXJlcik6CisgICAgICAgIChXZWJDb3JlOjpOb2RlUmVu ZGVyaW5nQ29udGV4dDo6cHJldmlvdXNSZW5kZXJlcik6CisgICAgICAgIChXZWJDb3JlOjpOb2Rl UmVuZGVyaW5nQ29udGV4dDo6cGFyZW50UmVuZGVyZXIpOgorICAgICAgICAoV2ViQ29yZTo6Tm9k ZVJlbmRlcmluZ0NvbnRleHQ6OnNob3VsZENyZWF0ZVJlbmRlcmVyKToKKyAgICAgICAgKFdlYkNv cmU6Ok5vZGVSZW5kZXJpbmdDb250ZXh0Ojppc09uRW5jYXBzdWxhdGlvbkJvdW5kYXJ5KToKKyAg ICAgICAgKiBkb20vTm9kZVJlbmRlcmluZ0NvbnRleHQuaDoKKyAgICAgICAgKE5vZGVSZW5kZXJp bmdDb250ZXh0KToKKyAgICAgICAgKFdlYkNvcmU6Ok5vZGVSZW5kZXJpbmdDb250ZXh0OjpwYXJl bnROb2RlRm9yUmVuZGVyaW5nQW5kU3R5bGUpOgorICAgICAgICAoV2ViQ29yZTo6Tm9kZVJlbmRl cmluZ0NvbnRleHQ6OnJlc2V0U3R5bGVJbmhlcml0YW5jZSk6CisgICAgICAgIChXZWJDb3JlOjpO b2RlUmVuZGVyaW5nQ29udGV4dDo6aW5zZXJ0aW9uUG9pbnQpOgorCiAyMDEyLTExLTIyICBLZW50 YXJvIEhhcmEgIDxoYXJha2VuQGNocm9taXVtLm9yZz4KIAogICAgICAgICBbVjhdIE1vdmUgV29y a2VyRXhlY3V0aW9uQ29udGV4dFByb3h5Ojppbml0aWFsaXplSWZOZWVkZWQoKSB0byBWOEluaXRp YWxpemVyCkluZGV4OiBTb3VyY2UvV2ViQ29yZS9kb20vTm9kZVJlbmRlcmluZ0NvbnRleHQuY3Bw Cj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT0KLS0tIFNvdXJjZS9XZWJDb3JlL2RvbS9Ob2RlUmVuZGVyaW5nQ29udGV4dC5j cHAJKHJldmlzaW9uIDEzNTY2OCkKKysrIFNvdXJjZS9XZWJDb3JlL2RvbS9Ob2RlUmVuZGVyaW5n Q29udGV4dC5jcHAJKHdvcmtpbmcgY29weSkKQEAgLTU2LDExICs1NiwxOSBAQCBOb2RlUmVuZGVy aW5nQ29udGV4dDo6Tm9kZVJlbmRlcmluZ0NvbnRlCiAgICAgOiBtX25vZGUobm9kZSkKICAgICAs IG1fcGFyZW50Rmxvd1JlbmRlcmVyKDApCiB7CisjaWYgIUVOQUJMRShTSEFET1dfRE9NKQorICAg IENvbnRhaW5lck5vZGUqIHBhcmVudCA9IG1fbm9kZS0+cGFyZW50T3JIb3N0Tm9kZSgpOworICAg IG1fcGFyZW50Tm9kZUZvclJlbmRlcmluZ0FuZFN0eWxlID0gKHBhcmVudCAmJiBwYXJlbnQtPmlz U2hhZG93Um9vdCgpKSA/IHRvU2hhZG93Um9vdChwYXJlbnQpLT5ob3N0KCkgOiBwYXJlbnQ7Cisj ZWxzZQogICAgIENvbXBvc2VkU2hhZG93VHJlZVdhbGtlcjo6ZmluZFBhcmVudChtX25vZGUsICZt X3BhcmVudERldGFpbHMpOworI2VuZGlmCiB9CiAKIE5vZGVSZW5kZXJpbmdDb250ZXh0OjpOb2Rl UmVuZGVyaW5nQ29udGV4dChOb2RlKiBub2RlLCBSZW5kZXJTdHlsZSogc3R5bGUpCiAgICAgOiBt X25vZGUobm9kZSkKKyNpZiAhRU5BQkxFKFNIQURPV19ET00pCisgICAgLCBtX3BhcmVudE5vZGVG b3JSZW5kZXJpbmdBbmRTdHlsZSgwKQorI2VuZGlmCiAgICAgLCBtX3N0eWxlKHN0eWxlKQogICAg ICwgbV9wYXJlbnRGbG93UmVuZGVyZXIoMCkKIHsKQEAgLTcwLDYgKzc4LDE0IEBAIE5vZGVSZW5k ZXJpbmdDb250ZXh0Ojp+Tm9kZVJlbmRlcmluZ0NvbnQKIHsKIH0KIAorc3RhdGljIGJvb2wgaXNW YWxpZE5vbkZsb3dUaHJlYWRSZW5kZXJlcihSZW5kZXJPYmplY3QqIHJlbmRlcmVyKQoreworICAg IGlmICghcmVuZGVyZXIpCisgICAgICAgIHJldHVybiBmYWxzZTsKKyAgICAvLyBEbyBub3QgcmV0 dXJuIGVsZW1lbnRzIHRoYXQgYXJlIGF0dGFjaGVkIHRvIGEgZGlmZmVyZW50IGZsb3ctdGhyZWFk LgorICAgIHJldHVybiByZW5kZXJlci0+c3R5bGUoKS0+Zmxvd1RocmVhZCgpLmlzRW1wdHkoKTsK K30KKwogUmVuZGVyT2JqZWN0KiBOb2RlUmVuZGVyaW5nQ29udGV4dDo6bmV4dFJlbmRlcmVyKCkg Y29uc3QKIHsKICAgICBpZiAoUmVuZGVyT2JqZWN0KiByZW5kZXJlciA9IG1fbm9kZS0+cmVuZGVy ZXIoKSkKQEAgLTgwLDE5ICs5NiwyMiBAQCBSZW5kZXJPYmplY3QqIE5vZGVSZW5kZXJpbmdDb250 ZXh0OjpuZXh0CiAKICAgICAvLyBBdm9pZCBhbiBPKE5eMikgcHJvYmxlbSB3aXRoIHRoaXMgZnVu Y3Rpb24gYnkgbm90IGNoZWNraW5nIGZvcgogICAgIC8vIG5leHRSZW5kZXJlcigpIHdoZW4gdGhl IHBhcmVudCBlbGVtZW50IGhhc24ndCBhdHRhY2hlZCB5ZXQuCi0gICAgaWYgKG1fcGFyZW50RGV0 YWlscy5ub2RlKCkgJiYgIW1fcGFyZW50RGV0YWlscy5ub2RlKCktPmF0dGFjaGVkKCkpCisgICAg Tm9kZSogcGFyZW50Tm9kZSA9IHBhcmVudE5vZGVGb3JSZW5kZXJpbmdBbmRTdHlsZSgpOworICAg IGlmICghcGFyZW50Tm9kZSB8fCAhcGFyZW50Tm9kZS0+YXR0YWNoZWQoKSkKICAgICAgICAgcmV0 dXJuIDA7CiAKKyNpZiAhRU5BQkxFKFNIQURPV19ET00pCisgICAgZm9yIChOb2RlKiBjdXJyZW50 ID0gbV9ub2RlLT5uZXh0U2libGluZygpOyBjdXJyZW50OyBjdXJyZW50ID0gY3VycmVudC0+bmV4 dFNpYmxpbmcoKSkgeworICAgICAgICBpZiAoaXNWYWxpZE5vbkZsb3dUaHJlYWRSZW5kZXJlcihj dXJyZW50LT5yZW5kZXJlcigpKSkKKyAgICAgICAgICAgIHJldHVybiBjdXJyZW50LT5yZW5kZXJl cigpOworICAgIH0KKyNlbHNlCiAgICAgQ29tcG9zZWRTaGFkb3dUcmVlV2Fsa2VyIHdhbGtlciht X25vZGUpOwogICAgIGZvciAod2Fsa2VyLm5leHRTaWJsaW5nKCk7IHdhbGtlci5nZXQoKTsgd2Fs a2VyLm5leHRTaWJsaW5nKCkpIHsKLSAgICAgICAgaWYgKFJlbmRlck9iamVjdCogcmVuZGVyZXIg PSB3YWxrZXIuZ2V0KCktPnJlbmRlcmVyKCkpIHsKLSAgICAgICAgICAgIC8vIERvIG5vdCByZXR1 cm4gZWxlbWVudHMgdGhhdCBhcmUgYXR0YWNoZWQgdG8gYSBkaWZmZXJlbnQgZmxvdy10aHJlYWQu Ci0gICAgICAgICAgICBpZiAocmVuZGVyZXItPnN0eWxlKCkgJiYgIXJlbmRlcmVyLT5zdHlsZSgp LT5mbG93VGhyZWFkKCkuaXNFbXB0eSgpKQotICAgICAgICAgICAgICAgIGNvbnRpbnVlOwotICAg ICAgICAgICAgcmV0dXJuIHJlbmRlcmVyOwotICAgICAgICB9CisgICAgICAgIGlmIChpc1ZhbGlk Tm9uRmxvd1RocmVhZFJlbmRlcmVyKHdhbGtlci5nZXQoKS0+cmVuZGVyZXIoKSkpCisgICAgICAg ICAgICByZXR1cm4gd2Fsa2VyLmdldCgpLT5yZW5kZXJlcigpOwogICAgIH0KLQorI2VuZGlmCiAg ICAgcmV0dXJuIDA7CiB9CiAKQEAgLTEwNiwxNyArMTI1LDE4IEBAIFJlbmRlck9iamVjdCogTm9k ZVJlbmRlcmluZ0NvbnRleHQ6OnByZXYKIAogICAgIC8vIEZJWE1FOiBXZSBzaG91bGQgaGF2ZSB0 aGUgc2FtZSBPKE5eMikgYXZvaWRhbmNlIGFzIG5leHRSZW5kZXJlciBkb2VzCiAgICAgLy8gaG93 ZXZlciwgd2hlbiBJIHRyaWVkIGFkZGluZyBpdCwgc2V2ZXJhbCB0ZXN0cyBmYWlsZWQuCi0KKyNp ZiAhRU5BQkxFKFNIQURPV19ET00pCisgICAgZm9yIChOb2RlKiBjdXJyZW50ID0gbV9ub2RlLT5w cmV2aW91c1NpYmxpbmcoKTsgY3VycmVudDsgY3VycmVudCA9IGN1cnJlbnQtPnByZXZpb3VzU2li bGluZygpKSB7CisgICAgICAgIGlmIChpc1ZhbGlkTm9uRmxvd1RocmVhZFJlbmRlcmVyKGN1cnJl bnQtPnJlbmRlcmVyKCkpKQorICAgICAgICAgICAgcmV0dXJuIGN1cnJlbnQtPnJlbmRlcmVyKCk7 CisgICAgfQorI2Vsc2UKICAgICBDb21wb3NlZFNoYWRvd1RyZWVXYWxrZXIgd2Fsa2VyKG1fbm9k ZSk7CiAgICAgZm9yICh3YWxrZXIucHJldmlvdXNTaWJsaW5nKCk7IHdhbGtlci5nZXQoKTsgd2Fs a2VyLnByZXZpb3VzU2libGluZygpKSB7Ci0gICAgICAgIGlmIChSZW5kZXJPYmplY3QqIHJlbmRl cmVyID0gd2Fsa2VyLmdldCgpLT5yZW5kZXJlcigpKSB7Ci0gICAgICAgICAgICAvLyBEbyBub3Qg cmV0dXJuIGVsZW1lbnRzIHRoYXQgYXJlIGF0dGFjaGVkIHRvIGEgZGlmZmVyZW50IGZsb3ctdGhy ZWFkLgotICAgICAgICAgICAgaWYgKHJlbmRlcmVyLT5zdHlsZSgpICYmICFyZW5kZXJlci0+c3R5 bGUoKS0+Zmxvd1RocmVhZCgpLmlzRW1wdHkoKSkKLSAgICAgICAgICAgICAgICBjb250aW51ZTsK LSAgICAgICAgICAgIHJldHVybiByZW5kZXJlcjsKLSAgICAgICAgfQorICAgICAgICBpZiAoaXNW YWxpZE5vbkZsb3dUaHJlYWRSZW5kZXJlcih3YWxrZXIuZ2V0KCktPnJlbmRlcmVyKCkpKQorICAg ICAgICAgICAgcmV0dXJuIHdhbGtlci5nZXQoKS0+cmVuZGVyZXIoKTsKICAgICB9Ci0KKyNlbmRp ZgogICAgIHJldHVybiAwOwogfQogCkBAIC0xMjYsMjIgKzE0NiwyNCBAQCBSZW5kZXJPYmplY3Qq IE5vZGVSZW5kZXJpbmdDb250ZXh0OjpwYXJlCiAgICAgICAgIHJldHVybiByZW5kZXJlci0+cGFy ZW50KCk7CiAgICAgaWYgKG1fcGFyZW50Rmxvd1JlbmRlcmVyKQogICAgICAgICByZXR1cm4gbV9w YXJlbnRGbG93UmVuZGVyZXI7Ci0KLSAgICByZXR1cm4gbV9wYXJlbnREZXRhaWxzLm5vZGUoKSA/ IG1fcGFyZW50RGV0YWlscy5ub2RlKCktPnJlbmRlcmVyKCkgOiAwOworICAgIGlmIChDb250YWlu ZXJOb2RlKiBwYXJlbnROb2RlID0gcGFyZW50Tm9kZUZvclJlbmRlcmluZ0FuZFN0eWxlKCkpCisg ICAgICAgIHJldHVybiBwYXJlbnROb2RlLT5yZW5kZXJlcigpOworICAgIHJldHVybiAwOwogfQog CiBib29sIE5vZGVSZW5kZXJpbmdDb250ZXh0OjpzaG91bGRDcmVhdGVSZW5kZXJlcigpIGNvbnN0 CiB7CiAgICAgaWYgKCFtX25vZGUtPmRvY3VtZW50KCktPnNob3VsZENyZWF0ZVJlbmRlcmVycygp KQogICAgICAgICByZXR1cm4gZmFsc2U7Ci0gICAgaWYgKCFtX3BhcmVudERldGFpbHMubm9kZSgp KQotICAgICAgICByZXR1cm4gZmFsc2U7CiAgICAgUmVuZGVyT2JqZWN0KiBwYXJlbnRSZW5kZXJl ciA9IHRoaXMtPnBhcmVudFJlbmRlcmVyKCk7CiAgICAgaWYgKCFwYXJlbnRSZW5kZXJlcikKICAg ICAgICAgcmV0dXJuIGZhbHNlOwogICAgIGlmICghcGFyZW50UmVuZGVyZXItPmNhbkhhdmVDaGls ZHJlbigpKQogICAgICAgICByZXR1cm4gZmFsc2U7Ci0gICAgaWYgKCFtX3BhcmVudERldGFpbHMu bm9kZSgpLT5jaGlsZFNob3VsZENyZWF0ZVJlbmRlcmVyKCp0aGlzKSkKKyAgICBDb250YWluZXJO b2RlKiBwYXJlbnROb2RlID0gcGFyZW50Tm9kZUZvclJlbmRlcmluZ0FuZFN0eWxlKCk7CisgICAg aWYgKCFwYXJlbnROb2RlKQorICAgICAgICByZXR1cm4gZmFsc2U7CisgICAgaWYgKCFwYXJlbnRO b2RlLT5jaGlsZFNob3VsZENyZWF0ZVJlbmRlcmVyKCp0aGlzKSkKICAgICAgICAgcmV0dXJuIGZh bHNlOwogICAgIHJldHVybiB0cnVlOwogfQpAQCAtMTc2LDcgKzE5OCwxMyBAQCB2b2lkIE5vZGVS ZW5kZXJpbmdDb250ZXh0Ojptb3ZlVG9GbG93VGhyCiAKIGJvb2wgTm9kZVJlbmRlcmluZ0NvbnRl eHQ6OmlzT25FbmNhcHN1bGF0aW9uQm91bmRhcnkoKSBjb25zdAogewotICAgIHJldHVybiBpc09u VXBwZXJFbmNhcHN1bGF0aW9uQm91bmRhcnkoKSB8fCBpc0xvd2VyRW5jYXBzdWxhdGlvbkJvdW5k YXJ5KG1fcGFyZW50RGV0YWlscy5pbnNlcnRpb25Qb2ludCgpKSB8fCBpc0xvd2VyRW5jYXBzdWxh dGlvbkJvdW5kYXJ5KG1fbm9kZS0+cGFyZW50Tm9kZSgpKTsKKyAgICBpZiAoaXNPblVwcGVyRW5j YXBzdWxhdGlvbkJvdW5kYXJ5KCkgfHwgaXNMb3dlckVuY2Fwc3VsYXRpb25Cb3VuZGFyeShtX25v ZGUtPnBhcmVudE5vZGUoKSkpCisgICAgICAgIHJldHVybiB0cnVlOworI2lmIEVOQUJMRShTSEFE T1dfRE9NKQorICAgIGlmIChpc0xvd2VyRW5jYXBzdWxhdGlvbkJvdW5kYXJ5KG1fcGFyZW50RGV0 YWlscy5pbnNlcnRpb25Qb2ludCgpKSkKKyAgICAgICAgcmV0dXJuIHRydWU7CisjZW5kaWYKKyAg ICByZXR1cm4gZmFsc2U7CiB9CiAKIGJvb2wgTm9kZVJlbmRlcmluZ0NvbnRleHQ6OmlzT25VcHBl ckVuY2Fwc3VsYXRpb25Cb3VuZGFyeSgpIGNvbnN0CkluZGV4OiBTb3VyY2UvV2ViQ29yZS9kb20v Tm9kZVJlbmRlcmluZ0NvbnRleHQuaAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2ViQ29yZS9kb20v Tm9kZVJlbmRlcmluZ0NvbnRleHQuaAkocmV2aXNpb24gMTM1NjY4KQorKysgU291cmNlL1dlYkNv cmUvZG9tL05vZGVSZW5kZXJpbmdDb250ZXh0LmgJKHdvcmtpbmcgY29weSkKQEAgLTY5LDcgKzY5 LDExIEBAIHByaXZhdGU6CiAgICAgdm9pZCBtb3ZlVG9GbG93VGhyZWFkSWZOZWVkZWQoKTsKIAog ICAgIE5vZGUqIG1fbm9kZTsKKyNpZiAhRU5BQkxFKFNIQURPV19ET00pCisgICAgQ29udGFpbmVy Tm9kZSogbV9wYXJlbnROb2RlRm9yUmVuZGVyaW5nQW5kU3R5bGU7CisjZWxzZQogICAgIENvbXBv c2VkU2hhZG93VHJlZVdhbGtlcjo6UGFyZW50VHJhdmVyc2FsRGV0YWlscyBtX3BhcmVudERldGFp bHM7CisjZW5kaWYKICAgICBSZWZQdHI8UmVuZGVyU3R5bGU+IG1fc3R5bGU7CiAgICAgUmVuZGVy TmFtZWRGbG93VGhyZWFkKiBtX3BhcmVudEZsb3dSZW5kZXJlcjsKICAgICBBdG9taWNTdHJpbmcg bV9mbG93VGhyZWFkOwpAQCAtODIsMTIgKzg2LDIwIEBAIGlubGluZSBOb2RlKiBOb2RlUmVuZGVy aW5nQ29udGV4dDo6bm9kZSgKIAogaW5saW5lIENvbnRhaW5lck5vZGUqIE5vZGVSZW5kZXJpbmdD b250ZXh0OjpwYXJlbnROb2RlRm9yUmVuZGVyaW5nQW5kU3R5bGUoKSBjb25zdAogeworI2lmICFF TkFCTEUoU0hBRE9XX0RPTSkKKyAgICByZXR1cm4gbV9wYXJlbnROb2RlRm9yUmVuZGVyaW5nQW5k U3R5bGU7CisjZWxzZQogICAgIHJldHVybiBtX3BhcmVudERldGFpbHMubm9kZSgpOworI2VuZGlm CiB9CiAKIGlubGluZSBib29sIE5vZGVSZW5kZXJpbmdDb250ZXh0OjpyZXNldFN0eWxlSW5oZXJp dGFuY2UoKSBjb25zdAogeworI2lmICFFTkFCTEUoU0hBRE9XX0RPTSkKKyAgICByZXR1cm4gZmFs c2U7CisjZWxzZQogICAgIHJldHVybiBtX3BhcmVudERldGFpbHMucmVzZXRTdHlsZUluaGVyaXRh bmNlKCk7CisjZW5kaWYKIH0KIAogaW5saW5lIGNvbnN0IFJlbmRlclN0eWxlKiBOb2RlUmVuZGVy aW5nQ29udGV4dDo6c3R5bGUoKSBjb25zdApAQCAtOTcsNyArMTA5LDExIEBAIGlubGluZSBjb25z dCBSZW5kZXJTdHlsZSogTm9kZVJlbmRlcmluZ0MKIAogaW5saW5lIEluc2VydGlvblBvaW50KiBO b2RlUmVuZGVyaW5nQ29udGV4dDo6aW5zZXJ0aW9uUG9pbnQoKSBjb25zdAogeworI2lmICFFTkFC TEUoU0hBRE9XX0RPTSkKKyAgICByZXR1cm4gMDsKKyNlbHNlCiAgICAgcmV0dXJuIG1fcGFyZW50 RGV0YWlscy5pbnNlcnRpb25Qb2ludCgpOworI2VuZGlmCiB9CiAKIH0gLy8gbmFtZXNwYWNlIFdl YkNvcmUK