111494 2013-03-05 16:30:19 -0800 innerHTML should use lazyAttach to avoid creating a render tree until needed 2023-04-01 00:55:11 -0700 1 1 1 Unclassified WebKit New Bugs 528+ (Nightly build) Unspecified Unspecified RESOLVED CONFIGURATION CHANGED P2 Normal --- 116926 94733 111644 1 eric eric abarth aestes annevk buildbot dglazkov esprehn haraken jchaffraix kling koivisto mifenton rniwa tonyg webkit.review.bot oldest_to_newest 848229 0 eric 2013-03-05 16:30:19 -0800 innerHTML should use lazyAttach to avoid creating a render tree until needed 848232 1 191596 eric 2013-03-05 16:33:36 -0800 Created attachment 191596 Patch 848235 2 eric 2013-03-05 16:34:17 -0800 This may need a couple more tweeks. I saw two editing-related failures on my local machine, but am not yet sure if those are from this patch or not. 848239 3 191596 haraken 2013-03-05 16:35:35 -0800 Comment on attachment 191596 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=191596&action=review > Source/WebCore/ChangeLog:15 > + In any case, this brings our inner-html-setter number from 93 runs/sec to 507 runs/sec :) Really! 848242 4 191596 abarth 2013-03-05 16:39:28 -0800 Comment on attachment 191596 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=191596&action=review > Source/WebCore/editing/markup.cpp:1100 > - containerNode->replaceChild(fragment, containerNode->firstChild(), ec); > + containerNode->replaceChild(fragment, containerNode->firstChild(), ec, true); I wish the last parameter had more readable value than "true" (but that's obviously not something you need to solve in this patch). 848243 5 haraken 2013-03-05 16:40:45 -0800 The 2400% regression bug is here: https://bugs.webkit.org/show_bug.cgi?id=94733 848244 6 eric 2013-03-05 16:41:07 -0800 (In reply to comment #4) > (From update of attachment 191596 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=191596&action=review > > > Source/WebCore/editing/markup.cpp:1100 > > - containerNode->replaceChild(fragment, containerNode->firstChild(), ec); > > + containerNode->replaceChild(fragment, containerNode->firstChild(), ec, true); > > I wish the last parameter had more readable value than "true" (but that's obviously not something you need to solve in this patch). I tried shaving that Yak first, actually. :) But that will wait for a second patch. I'm going to also change the default from false to true, since I suspect any place which depends on implicit "false" behavior is probably wrong. There are some cases where it might make performance-sense to explicitly require a synchronous attach (like for editing I could imagine). 848310 7 eric 2013-03-05 17:43:41 -0800 (In reply to comment #4) > (From update of attachment 191596 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=191596&action=review > > > Source/WebCore/editing/markup.cpp:1100 > > - containerNode->replaceChild(fragment, containerNode->firstChild(), ec); > > + containerNode->replaceChild(fragment, containerNode->firstChild(), ec, true); > > I wish the last parameter had more readable value than "true" (but that's obviously not something you need to solve in this patch). Bug 111503. 848320 8 eric 2013-03-05 17:54:43 -0800 If bug 111503 lands before this one, this will need a trivial rebase. 848396 9 191596 webkit.review.bot 2013-03-05 19:46:52 -0800 Comment on attachment 191596 Patch Attachment 191596 did not pass chromium-ews (chromium-xvfb): Output: http://webkit-commit-queue.appspot.com/results/17082077 New failing tests: fast/forms/basic-buttons.html editing/execCommand/change-font.html 848568 10 191596 buildbot 2013-03-06 00:42:49 -0800 Comment on attachment 191596 Patch Attachment 191596 did not pass mac-wk2-ews (mac-wk2): Output: http://webkit-commit-queue.appspot.com/results/17081088 848675 11 191596 buildbot 2013-03-06 03:09:20 -0800 Comment on attachment 191596 Patch Attachment 191596 did not pass win-ews (win): Output: http://webkit-commit-queue.appspot.com/results/17021095 848829 12 191596 buildbot 2013-03-06 07:04:32 -0800 Comment on attachment 191596 Patch Attachment 191596 did not pass mac-ews (mac): Output: http://webkit-commit-queue.appspot.com/results/17065088 849167 13 eric 2013-03-06 12:07:59 -0800 --- /src/WebKit/Source/WebKit/chromium/webkit/Release/layout-test-results/fast/forms/basic-buttons-expected.txt +++ /src/WebKit/Source/WebKit/chromium/webkit/Release/layout-test-results/fast/forms/basic-buttons-actual.txt @@ -14,6 +14,7 @@ text run at (0,40) width 617: "with text (\"foo\") and then uses six different paddings to make sure each of the buttons render properly. " RenderBR {BR} at (617,40) size 0x19 RenderBR {BR} at (0,60) size 0x19 + RenderText {#text} at (0,0) size 0x0 RenderTable {TABLE} at (0,80) size 657x285 RenderTableSection {TBODY} at (0,0) size 657x285 RenderTableRow {TR} at (0,0) size 657x22 And: --- /src/WebKit/Source/WebKit/chromium/webkit/Release/layout-test-results/editing/execCommand/change-font-expected.txt +++ /src/WebKit/Source/WebKit/chromium/webkit/Release/layout-test-results/editing/execCommand/change-font-actual.txt @@ -1,2 +1,5 @@ There should only be one font tag. SUCCESS + + Are the failures. Both come from uses of innerHTML on the body element. I'm not sure about either. Maybe there is some white-space collapsing that we would normally do during synchronous attach() which we somehow miss in the async case? The rendering of both pages is correct, this is a DRT-only issue. Not yet sure if we care. :) 849198 14 eric 2013-03-06 12:41:22 -0800 These failures sound very related to the code Haraken touched in bug 110786. (I'm not suggesting that his change broke anything here. Just that he's touched that code and it's probably not lazy-attach aware.) 849203 15 eric 2013-03-06 12:45:52 -0800 (In reply to comment #14) > These failures sound very related to the code Haraken touched in bug 110786. (I'm not suggesting that his change broke anything here. Just that he's touched that code and it's probably not lazy-attach aware.) I suspect that we should move the "don't ignore text renderers in all cases" code in Node::attach() to style recalc and that would fix this issue. Is suspect we're currently broken wrt ignoring text nodes and display/visibility changes of renderers. 849211 16 eric 2013-03-06 12:52:39 -0800 This FIXME in Element::recalcStyle *terrifies* me: if (ch == Detach || !currentStyle) { // FIXME: The style gets computed twice by calling attach. We could do better if we passed the style along. reattach(); If we're really recalc-ing style twice for intial attach of a lazy-attached node, that's absolutely horrible. 849316 17 eric 2013-03-06 14:41:48 -0800 (In reply to comment #16) > This FIXME in Element::recalcStyle *terrifies* me: > if (ch == Detach || !currentStyle) { > // FIXME: The style gets computed twice by calling attach. We could do better if we passed the style along. > reattach(); > > If we're really recalc-ing style twice for intial attach of a lazy-attached node, that's absolutely horrible. Filed bug 111627. 849998 18 191596 koivisto 2013-03-07 05:41:57 -0800 Comment on attachment 191596 Patch We should really do all attaching lazily (including that triggered by the parser). 850412 19 eric 2013-03-07 13:03:50 -0800 (In reply to comment #18) > (From update of attachment 191596 [details]) > We should really do all attaching lazily (including that triggered by the parser). Agreed. Do you have any theories for the failures mentioned in comment 13? Debugging those is blocking this change. 850962 20 koivisto 2013-03-08 00:42:38 -0800 > Agreed. Do you have any theories for the failures mentioned in comment 13? Not without debugging no. 854794 21 eric 2013-03-13 16:59:02 -0700 http://trac.webkit.org/browser/trunk/LayoutTests/fast/forms/basic-buttons.html and http://trac.webkit.org/browser/trunk/LayoutTests/editing/execCommand/change-font.html are the failing tests. 854805 22 193025 eric 2013-03-13 17:14:15 -0700 Created attachment 193025 updated to tot 887626 23 eric 2013-05-10 14:41:56 -0700 With the patch applied, this test: <!DOCTYPE html> <html> <body> text <br> <script> document.body.innerHTML += '<div></div>'; </script> </body> </html> Fails with this diff: @@ -7,4 +7,7 @@ RenderText {#text} at (0,0) size 26x19 text run at (0,0) width 26: "text " RenderBR {BR} at (26,0) size 0x19 + RenderText {#text} at (0,0) size 0x0 RenderBlock {DIV} at (0,20) size 784x0 + RenderBlock (anonymous) at (0,20) size 784x0 + RenderText {#text} at (0,0) size 0x0 Still debuggin. 887644 24 eric 2013-05-10 15:27:36 -0700 Interestingly, removing this block in Node::attach() makes the behavior change go away. :) (At least in the test in question.) // If this node got a renderer it may be the previousRenderer() of sibling text nodes and thus affect the // result of Text::textRendererIsNeeded() for those nodes. if (renderer()) { for (Node* next = nextSibling(); next; next = next->nextSibling()) { if (next->renderer()) break; if (!next->attached()) break; // Assume this means none of the following siblings are attached. if (!next->isTextNode()) continue; ASSERT(!next->renderer()); toText(next)->createTextRendererIfNeeded(); // If we again decided not to create a renderer for next, we can bail out the loop, // because it won't affect the result of Text::textRendererIsNeeded() for the rest // of sibling nodes. if (!next->renderer()) break; } } I'm not yet entirely sure what the "correct" behavior for these tests are, or if that block is doing what it really means to. 887686 25 eric 2013-05-10 17:24:44 -0700 The following printf-logging on the above test: @@ -221,8 +221,11 @@ bool Text::textRendererIsNeeded(const NodeRenderingContext& context) return true; RenderObject* prev = context.previousRenderer(); - if (prev && prev->isBR()) // <span><br/> <br/></span> + printf("prev: %p, br: %i\n", prev, prev && prev->isBR()); + if (prev && prev->isBR()) { + // <span><br/> <br/></span> return false; + } if (parent->isRenderInline()) { // <span><div/> <div/></span> @@ -236,12 +239,15 @@ bool Text::textRendererIsNeeded(const NodeRenderingContext& context) while (first && first->isFloatingOrOutOfFlowPositioned()) first = first->nextSibling(); RenderObject* next = context.nextRenderer(); - if (!first || next == first) + if (!first || next == first) { + printf("magic!\n"); // Whitespace at the start of a block just goes away. Don't even // make a render object for this text. return false; } + } + printf("Default true: '%s'\n", wholeText().utf8().data()); return true; } Produces the following change: @@ -1,8 +1,19 @@ -prev: 0x3c7cbc586038, br: 1 -prev: 0x3c7cbc3594b8, br: 1 -prev: 0x3c7cbc6dc138, br: 0 -prev: 0x3c7cbc6dc138, br: 0 -prev: 0x3c7cbc6dc138, br: 0 +prev: 0x1af39d059578, br: 1 +prev: (nil), br: 0 +magic! +prev: (nil), br: 0 +magic! +prev: (nil), br: 0 +magic! +prev: 0x1af39d4ee0f8, br: 0 +Default true: ' +' +prev: 0x1af39d512cf8, br: 0 +Default true: ' + + + +' layer at (0,0) size 800x600 RenderView at (0,0) size 800x600 layer at (0,0) size 800x36 @@ -12,4 +23,7 @@ RenderText {#text} at (0,0) size 26x19 text run at (0,0) width 26: "text " RenderBR {BR} at (26,0) size 0x19 + RenderText {#text} at (0,0) size 0x0 RenderBlock {DIV} at (0,20) size 784x0 + RenderBlock (anonymous) at (0,20) size 784x0 + RenderText {#text} at (0,0) size 0x0 Clearly we're taking a very different path for text node creation during lazyAttach. Still investigating. 887693 26 rniwa 2013-05-10 17:40:56 -0700 (In reply to comment #24) > Interestingly, removing this block in Node::attach() makes the behavior change go away. :) (At least in the test in question.) How about reverting http://trac.webkit.org/changeset/144526 ? 887723 27 eric 2013-05-10 18:31:53 -0700 I believe the problem here is simply the code in Node::attach(). When lazy-attaching, nothing has a renderer yet. But the first thing we attach, then walks through its siblings, creating a text renderers. (See comment 24 for the context.) lazyAttach marks you as "attached", which confuses this code. 1945931 28 annevk 2023-04-01 00:55:11 -0700 This has all been substantially changed. 191596 2013-03-05 16:33:36 -0800 2013-03-13 17:14:10 -0700 Patch bug-111494-20130305162952.patch text/plain 2732 eric U3VidmVyc2lvbiBSZXZpc2lvbjogMTQ0ODAxCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggMzI3ODYyNDhkNDk3ZDdi Njg1OWYyYWFiOGYxZDcyOGM3ZmFhYmVhMC4uYjhiOGI2MmEyYTU2NThlZGJhNmVjNjAxZTAxODhm OTJmMmUwMDQ1MiAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1Nv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDIzIEBACisyMDEzLTAzLTA1ICBFcmlj IFNlaWRlbCAgPGVyaWNAd2Via2l0Lm9yZz4KKworICAgICAgICBpbm5lckhUTUwgc2hvdWxkIHVz ZSBsYXp5QXR0YWNoIHRvIGF2b2lkIGNyZWF0aW5nIGEgcmVuZGVyIHRyZWUgdW50aWwgbmVlZGVk CisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xMTE0OTQK KworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBJdCB3YXMg YnJvdWdodCB0byBteSBhdHRlbnRpb24gdGhhdCBpbm5lckhUTUwgaXMgbXVjaCBzbG93ZXIgKDI0 MDAlIGknbSB0b2xkISkKKyAgICAgICAgdGhhbiBiZWZvcmUgdGhlIGh0bWw1IHBhcnNlci4KKyAg ICAgICAgTG9va2luZyBhdCBQZXJmb3JtYW5jZVRlc3RzL1BhcnNlci9pbm5lci1odG1sLXNldHRl ci5odG1sIEkgc2VlIHRoZSBzYW1wbGUKKyAgICAgICAgc3BlbmRpbmcgbXVjaCBvZiBpdHMgdGlt ZSBpbiBjcmVhdGluZyByZW5kZXJlcnMsIHdoaWNoIEkgZG9uJ3QgYmVsaWV2ZSBpdCB1c2VkIHRv IGRvLgorICAgICAgICBUaGlzIG1heSBoYXZlIGNvbWUgZnJvbSBodHRwOi8vdHJhYy53ZWJraXQu b3JnL2NoYW5nZXNldC82NTM3MgorICAgICAgICBvciBtb3JlIHJlY2VudCBTaGFkb3dET00gY2hh bmdlcywgaXQncyB1bmNsZWFyLgorCisgICAgICAgIEluIGFueSBjYXNlLCB0aGlzIGJyaW5ncyBv dXIgaW5uZXItaHRtbC1zZXR0ZXIgbnVtYmVyIGZyb20gOTMgcnVucy9zZWMgdG8gNTA3IHJ1bnMv c2VjIDopCisKKyAgICAgICAgKiBlZGl0aW5nL21hcmt1cC5jcHA6CisgICAgICAgIChXZWJDb3Jl OjpyZXBsYWNlQ2hpbGRyZW5XaXRoRnJhZ21lbnQpOgorICAgICAgICAoV2ViQ29yZTo6cmVwbGFj ZUNoaWxkcmVuV2l0aFRleHQpOgorCiAyMDEzLTAzLTA1ICBUb255IEdlbnRpbGNvcmUgIDx0b255 Z0BjaHJvbWl1bS5vcmc+CiAKICAgICAgICAgQ29tcGFjdEhUTUxUb2tlbiBkb2VzIG5vdCBuZWVk IGFuIFhTU0luZm8gcG9pbnRlcgpkaWZmIC0tZ2l0IGEvU291cmNlL1dlYkNvcmUvZWRpdGluZy9t YXJrdXAuY3BwIGIvU291cmNlL1dlYkNvcmUvZWRpdGluZy9tYXJrdXAuY3BwCmluZGV4IGQzMjc3 YzM2NjY2NjdhM2RhNjI5MmE3YTc5YTUxN2VlMTU0MTJjNWMuLmUzOTJjODA0MjFkYWUzZTYxMDY3 NGY1ZDJlNzBmMzNhZjJhMjQyMWQgMTAwNjQ0Ci0tLSBhL1NvdXJjZS9XZWJDb3JlL2VkaXRpbmcv bWFya3VwLmNwcAorKysgYi9Tb3VyY2UvV2ViQ29yZS9lZGl0aW5nL21hcmt1cC5jcHAKQEAgLTEw OTcsMTIgKzEwOTcsMTIgQEAgdm9pZCByZXBsYWNlQ2hpbGRyZW5XaXRoRnJhZ21lbnQoQ29udGFp bmVyTm9kZSogY29udGFpbmVyLCBQYXNzUmVmUHRyPERvY3VtZW50RnIKICAgICB9CiAKICAgICBp ZiAoaGFzT25lQ2hpbGQoY29udGFpbmVyTm9kZS5nZXQoKSkpIHsKLSAgICAgICAgY29udGFpbmVy Tm9kZS0+cmVwbGFjZUNoaWxkKGZyYWdtZW50LCBjb250YWluZXJOb2RlLT5maXJzdENoaWxkKCks IGVjKTsKKyAgICAgICAgY29udGFpbmVyTm9kZS0+cmVwbGFjZUNoaWxkKGZyYWdtZW50LCBjb250 YWluZXJOb2RlLT5maXJzdENoaWxkKCksIGVjLCB0cnVlKTsKICAgICAgICAgcmV0dXJuOwogICAg IH0KIAogICAgIGNvbnRhaW5lck5vZGUtPnJlbW92ZUNoaWxkcmVuKCk7Ci0gICAgY29udGFpbmVy Tm9kZS0+YXBwZW5kQ2hpbGQoZnJhZ21lbnQsIGVjKTsKKyAgICBjb250YWluZXJOb2RlLT5hcHBl bmRDaGlsZChmcmFnbWVudCwgZWMsIHRydWUpOwogfQogCiB2b2lkIHJlcGxhY2VDaGlsZHJlbldp dGhUZXh0KENvbnRhaW5lck5vZGUqIGNvbnRhaW5lciwgY29uc3QgU3RyaW5nJiB0ZXh0LCBFeGNl cHRpb25Db2RlJiBlYykKQEAgLTExMTksMTIgKzExMTksMTIgQEAgdm9pZCByZXBsYWNlQ2hpbGRy ZW5XaXRoVGV4dChDb250YWluZXJOb2RlKiBjb250YWluZXIsIGNvbnN0IFN0cmluZyYgdGV4dCwg RXhjZXAKICAgICBSZWZQdHI8VGV4dD4gdGV4dE5vZGUgPSBUZXh0OjpjcmVhdGUoY29udGFpbmVy Tm9kZS0+ZG9jdW1lbnQoKSwgdGV4dCk7CiAKICAgICBpZiAoaGFzT25lQ2hpbGQoY29udGFpbmVy Tm9kZS5nZXQoKSkpIHsKLSAgICAgICAgY29udGFpbmVyTm9kZS0+cmVwbGFjZUNoaWxkKHRleHRO b2RlLnJlbGVhc2UoKSwgY29udGFpbmVyTm9kZS0+Zmlyc3RDaGlsZCgpLCBlYyk7CisgICAgICAg IGNvbnRhaW5lck5vZGUtPnJlcGxhY2VDaGlsZCh0ZXh0Tm9kZS5yZWxlYXNlKCksIGNvbnRhaW5l ck5vZGUtPmZpcnN0Q2hpbGQoKSwgZWMsIHRydWUpOwogICAgICAgICByZXR1cm47CiAgICAgfQog CiAgICAgY29udGFpbmVyTm9kZS0+cmVtb3ZlQ2hpbGRyZW4oKTsKLSAgICBjb250YWluZXJOb2Rl LT5hcHBlbmRDaGlsZCh0ZXh0Tm9kZS5yZWxlYXNlKCksIGVjKTsKKyAgICBjb250YWluZXJOb2Rl LT5hcHBlbmRDaGlsZCh0ZXh0Tm9kZS5yZWxlYXNlKCksIGVjLCB0cnVlKTsKIH0KIAogfQo= 193025 2013-03-13 17:14:15 -0700 2013-03-13 17:14:15 -0700 updated to tot bug-111494-20130313171012.patch text/plain 2808 eric U3VidmVyc2lvbiBSZXZpc2lvbjogMTQ1NzcyCmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggYTQxZGRlYTE3YmNiMjk4 NjJjYzlmYzdjOGEyOGJkYWU3NGJlOWViZC4uN2FjMDYwMzU2ZjAxODMwYmMxODM4NDJkYmY2N2E4 MTQ0NmZmMjFlNyAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1Nv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDIzIEBACisyMDEzLTAzLTEzICBFcmlj IFNlaWRlbCAgPGVyaWNAd2Via2l0Lm9yZz4KKworICAgICAgICBpbm5lckhUTUwgc2hvdWxkIHVz ZSBsYXp5QXR0YWNoIHRvIGF2b2lkIGNyZWF0aW5nIGEgcmVuZGVyIHRyZWUgdW50aWwgbmVlZGVk CisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0xMTE0OTQK KworICAgICAgICBSZXZpZXdlZCBieSBEaW1pdHJpIEdsYXprb3YuCisKKyAgICAgICAgSXQgd2Fz IGJyb3VnaHQgdG8gbXkgYXR0ZW50aW9uIHRoYXQgaW5uZXJIVE1MIGlzIG11Y2ggc2xvd2VyICgy NDAwJSBpJ20gdG9sZCEpCisgICAgICAgIHRoYW4gYmVmb3JlIHRoZSBodG1sNSBwYXJzZXIuCisg ICAgICAgIExvb2tpbmcgYXQgUGVyZm9ybWFuY2VUZXN0cy9QYXJzZXIvaW5uZXItaHRtbC1zZXR0 ZXIuaHRtbCBJIHNlZSB0aGUgc2FtcGxlCisgICAgICAgIHNwZW5kaW5nIG11Y2ggb2YgaXRzIHRp bWUgaW4gY3JlYXRpbmcgcmVuZGVyZXJzLCB3aGljaCBJIGRvbid0IGJlbGlldmUgaXQgdXNlZCB0 byBkby4KKyAgICAgICAgVGhpcyBtYXkgaGF2ZSBjb21lIGZyb20gaHR0cDovL3RyYWMud2Via2l0 Lm9yZy9jaGFuZ2VzZXQvNjUzNzIKKyAgICAgICAgb3IgbW9yZSByZWNlbnQgU2hhZG93RE9NIGNo YW5nZXMsIGl0J3MgdW5jbGVhci4KKworICAgICAgICBJbiBhbnkgY2FzZSwgdGhpcyBicmluZ3Mg b3VyIGlubmVyLWh0bWwtc2V0dGVyIG51bWJlciBmcm9tIDkzIHJ1bnMvc2VjIHRvIDUwNyBydW5z L3NlYyA6KQorCisgICAgICAgICogZWRpdGluZy9tYXJrdXAuY3BwOgorICAgICAgICAoV2ViQ29y ZTo6cmVwbGFjZUNoaWxkcmVuV2l0aEZyYWdtZW50KToKKyAgICAgICAgKFdlYkNvcmU6OnJlcGxh Y2VDaGlsZHJlbldpdGhUZXh0KToKKwogMjAxMy0wMy0xMyAgUmFmYWVsIFdlaW5zdGVpbiAgPHJh ZmFlbHdAY2hyb21pdW0ub3JnPgogCiAgICAgICAgIFtIVE1MVGVtcGxhdGVFbGVtZW50XSBFT0Yg c2hvdWxkIHVud2luZCB0aGUgc3RhY2sgb2YgdGVtcGxhdGUgaW5zZXJ0aW9uIG1vZGVzIGl0ZXJh dGl2ZWx5CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9lZGl0aW5nL21hcmt1cC5jcHAgYi9T b3VyY2UvV2ViQ29yZS9lZGl0aW5nL21hcmt1cC5jcHAKaW5kZXggNmEzYjJhMDY0NWI5OTk2MDNk ZTgyZDJiMWYzZmMyOTA3YmZkYWQ3NC4uMTY3ZTU3YzFiMTk2ZDM4N2E5NWNmYmFkMzVmMDA2NDU1 MzBiMzcxMiAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvZWRpdGluZy9tYXJrdXAuY3BwCisr KyBiL1NvdXJjZS9XZWJDb3JlL2VkaXRpbmcvbWFya3VwLmNwcApAQCAtMTEwNiwxMiArMTEwNiwx MiBAQCB2b2lkIHJlcGxhY2VDaGlsZHJlbldpdGhGcmFnbWVudChDb250YWluZXJOb2RlKiBjb250 YWluZXIsIFBhc3NSZWZQdHI8RG9jdW1lbnRGcgogICAgIH0KIAogICAgIGlmIChoYXNPbmVDaGls ZChjb250YWluZXJOb2RlLmdldCgpKSkgewotICAgICAgICBjb250YWluZXJOb2RlLT5yZXBsYWNl Q2hpbGQoZnJhZ21lbnQsIGNvbnRhaW5lck5vZGUtPmZpcnN0Q2hpbGQoKSwgZWMpOworICAgICAg ICBjb250YWluZXJOb2RlLT5yZXBsYWNlQ2hpbGQoZnJhZ21lbnQsIGNvbnRhaW5lck5vZGUtPmZp cnN0Q2hpbGQoKSwgZWMsIEF0dGFjaExhemlseSk7CiAgICAgICAgIHJldHVybjsKICAgICB9CiAK ICAgICBjb250YWluZXJOb2RlLT5yZW1vdmVDaGlsZHJlbigpOwotICAgIGNvbnRhaW5lck5vZGUt PmFwcGVuZENoaWxkKGZyYWdtZW50LCBlYyk7CisgICAgY29udGFpbmVyTm9kZS0+YXBwZW5kQ2hp bGQoZnJhZ21lbnQsIGVjLCBBdHRhY2hMYXppbHkpOwogfQogCiB2b2lkIHJlcGxhY2VDaGlsZHJl bldpdGhUZXh0KENvbnRhaW5lck5vZGUqIGNvbnRhaW5lciwgY29uc3QgU3RyaW5nJiB0ZXh0LCBF eGNlcHRpb25Db2RlJiBlYykKQEAgLTExMjgsMTIgKzExMjgsMTIgQEAgdm9pZCByZXBsYWNlQ2hp bGRyZW5XaXRoVGV4dChDb250YWluZXJOb2RlKiBjb250YWluZXIsIGNvbnN0IFN0cmluZyYgdGV4 dCwgRXhjZXAKICAgICBSZWZQdHI8VGV4dD4gdGV4dE5vZGUgPSBUZXh0OjpjcmVhdGUoY29udGFp bmVyTm9kZS0+ZG9jdW1lbnQoKSwgdGV4dCk7CiAKICAgICBpZiAoaGFzT25lQ2hpbGQoY29udGFp bmVyTm9kZS5nZXQoKSkpIHsKLSAgICAgICAgY29udGFpbmVyTm9kZS0+cmVwbGFjZUNoaWxkKHRl eHROb2RlLnJlbGVhc2UoKSwgY29udGFpbmVyTm9kZS0+Zmlyc3RDaGlsZCgpLCBlYyk7CisgICAg ICAgIGNvbnRhaW5lck5vZGUtPnJlcGxhY2VDaGlsZCh0ZXh0Tm9kZS5yZWxlYXNlKCksIGNvbnRh aW5lck5vZGUtPmZpcnN0Q2hpbGQoKSwgZWMsIEF0dGFjaExhemlseSk7CiAgICAgICAgIHJldHVy bjsKICAgICB9CiAKICAgICBjb250YWluZXJOb2RlLT5yZW1vdmVDaGlsZHJlbigpOwotICAgIGNv bnRhaW5lck5vZGUtPmFwcGVuZENoaWxkKHRleHROb2RlLnJlbGVhc2UoKSwgZWMpOworICAgIGNv bnRhaW5lck5vZGUtPmFwcGVuZENoaWxkKHRleHROb2RlLnJlbGVhc2UoKSwgZWMsIEF0dGFjaExh emlseSk7CiB9CiAKIH0K