228780 2021-08-04 09:17:23 -0700 Tweak HTML parser yielding 2023-03-29 15:22:53 -0700 1 1 1 Unclassified WebKit WebKit Misc. WebKit Nightly Build Unspecified Unspecified RESOLVED FIXED https://bugs.webkit.org/show_bug.cgi?id=254637 InRadar P2 Normal --- 251943 1 pvollan pvollan aestes ap bfulgham cdumez changseok commit-queue darin esprehn+autocc ews-watchlist ggaren gyuyoung.kim koivisto mjs nham slewis webkit-bug-importer zalan oldest_to_newest 1781899 0 pvollan 2021-08-04 09:17:23 -0700 Tweak html parser yielding to attempt to improve page load performance with respect to certain page load metrics. 1781900 1 webkit-bug-importer 2021-08-04 09:17:47 -0700 <rdar://problem/81517847> 1781905 2 434910 pvollan 2021-08-04 09:41:15 -0700 Created attachment 434910 Patch 1781910 3 434910 bfulgham 2021-08-04 09:49:18 -0700 Comment on attachment 434910 Patch View in context: https://bugs.webkit.org/attachment.cgi?id=434910&action=review > Source/WebCore/html/parser/HTMLParserScheduler.cpp:-126 > - if (elapsedTime < elapsedTimeLimit) Is there any reason to not retain the 16 ms upper bound on yielding? I think with this change we will quickly yield if we have a layout pending after first meaningful paint, but if that's not true we might no longer yield at 16 ms. I guess I'm suggesting we retain the current time limit, but check for it after first checking for the pending layout. 1781911 4 434910 ggaren 2021-08-04 09:51:12 -0700 Comment on attachment 434910 Patch I think Antti needs to review this, since he recently established this behavior based on extensive testing, and there were lots of tradeoffs involved. Can you share which page load metric(s) improve when we make this change? I see the logic of not yielding when there is no pending layout (the main goal of yielding is to show something new to the user, and if there is no pending layout, there is nothing new to show); but why also test isVisuallyNonEmpty()? If we have not painted yet, and we have nothing to paint now, what is the point of yielding the parser? Wouldn't it be better to keep parsing, in the hopes of finding something new to show to the user? Side note: I believe the document->view()->isVisuallyNonEmpty() check tests for first paint, not first *meaningful* paint. I'm not sure if this influences the code at all, but assuming my understanding is correct, let's fix up the ChangeLog. 1781913 5 ggaren 2021-08-04 09:53:48 -0700 Also: Please see my comment in Radar. 1781924 6 zalan 2021-08-04 10:28:34 -0700 isVisuallyNonEmpty() does not really have much to do with _painting_ (it's a legacy term, we should probably change it). It only tells you whether we believe we've got enough content to present to the user. >document->view()->isVisuallyNonEmpty() && !document->isLayoutTimerActive() Is this about letting us paint sooner after we established that there's enough content to show (i.e. we are in between (a)figured we've got enough content to show (b)paint)? -If so, I am a bit confused by the layout timer check. 1781928 7 zalan 2021-08-04 10:38:22 -0700 >isVisuallyNonEmpty() does not really have much to do with _painting_ Let's make it a bit more accurate. The subsequent paint is supposed to be visually non-empty. 1782025 8 pvollan 2021-08-04 13:42:35 -0700 (In reply to Brent Fulgham from comment #3) > Comment on attachment 434910 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=434910&action=review > > > Source/WebCore/html/parser/HTMLParserScheduler.cpp:-126 > > - if (elapsedTime < elapsedTimeLimit) > > Is there any reason to not retain the 16 ms upper bound on yielding? I think > with this change we will quickly yield if we have a layout pending after > first meaningful paint, but if that's not true we might no longer yield at > 16 ms. > > I guess I'm suggesting we retain the current time limit, but check for it > after first checking for the pending layout. That is a good point. I replaced the absolute timeout since I think it can lead to difference in behavior between devices. I will try measuring performance with your suggested change. Thanks for reviewing! 1782031 9 pvollan 2021-08-04 13:46:04 -0700 (In reply to Geoffrey Garen from comment #4) > Comment on attachment 434910 [details] > Patch > > I think Antti needs to review this, since he recently established this > behavior based on extensive testing, and there were lots of tradeoffs > involved. > > Can you share which page load metric(s) improve when we make this change? > See radar for more information on this. > I see the logic of not yielding when there is no pending layout (the main > goal of yielding is to show something new to the user, and if there is no > pending layout, there is nothing new to show); but why also test > isVisuallyNonEmpty()? If we have not painted yet, and we have nothing to > paint now, what is the point of yielding the parser? Wouldn't it be better > to keep parsing, in the hopes of finding something new to show to the user? > Yes, that is a good point. I will try measuring performance without the check for isVisuallyNonEmpty(). > Side note: I believe the document->view()->isVisuallyNonEmpty() check tests > for first paint, not first *meaningful* paint. I'm not sure if this > influences the code at all, but assuming my understanding is correct, let's > fix up the ChangeLog. Ah, that is good to know. Thanks for reviewing! 1782034 10 pvollan 2021-08-04 13:58:15 -0700 (In reply to zalan from comment #6) > isVisuallyNonEmpty() does not really have much to do with _painting_ (it's a > legacy term, we should probably change it). It only tells you whether we > believe we've got enough content to present to the user. > Ah, thanks, that is good to know! > >document->view()->isVisuallyNonEmpty() && !document->isLayoutTimerActive() > Is this about letting us paint sooner after we established that there's > enough content to show (i.e. we are in between (a)figured we've got enough > content to show (b)paint)? -If so, I am a bit confused by the layout timer > check. The intention behind this was to yield if there is a pending layout in order to give layout and painting priority. Although, as Geoff pointed out, the check for isVisuallyNonEmpty() might be incorrect, then. Thanks for reviewing! 1782177 11 434910 koivisto 2021-08-05 05:31:34 -0700 Comment on attachment 434910 Patch ASi comp PLT is showing 2.5% regression from this patch. No data from intel bots. 1782231 12 pvollan 2021-08-05 09:13:19 -0700 (In reply to Antti Koivisto from comment #11) > Comment on attachment 434910 [details] > Patch > > ASi comp PLT is showing 2.5% regression from this patch. No data from intel > bots. I will look into if further tuning is possible to avoid this regression while keeping the progression. Thanks for reviewing! 1782248 13 ggaren 2021-08-05 10:03:05 -0700 I'd be interested in an A/B test of just adding if (!document->isLayoutTimerActive()) return false; In theory, this might improve Competitive PLT by reducing parser yielding when we have nothing to render. But it might harm PLT5's domContentLoaded metric. I'd also be interested in an A/B test of adding if (!document->isLayoutTimerActive()) return false; and removing if (elapsedTime < elapsedTimeLimit) return false; I believe the check of isVisuallyNonEmpty() is harmful to Competitive PLT (it seems to say "If I have enough content to render, then don't!"). So maybe excluding that check would improve this patch. 1782253 14 pvollan 2021-08-05 10:09:33 -0700 (In reply to Geoffrey Garen from comment #13) > I'd be interested in an A/B test of just adding > > if (!document->isLayoutTimerActive()) > return false; > > In theory, this might improve Competitive PLT by reducing parser yielding > when we have nothing to render. But it might harm PLT5's domContentLoaded > metric. > > > I'd also be interested in an A/B test of adding > > if (!document->isLayoutTimerActive()) > return false; > > and removing > > if (elapsedTime < elapsedTimeLimit) > return false; > > I believe the check of isVisuallyNonEmpty() is harmful to Competitive PLT > (it seems to say "If I have enough content to render, then don't!"). So > maybe excluding that check would improve this patch. Thanks, these are great suggestions :) I will be creating A/B tests for this. 1783212 15 435271 pvollan 2021-08-10 10:36:00 -0700 Created attachment 435271 Patch 1785206 16 435793 pvollan 2021-08-18 13:50:17 -0700 Created attachment 435793 Patch 1929812 17 pvollan 2023-01-31 11:17:48 -0800 Pull request: https://github.com/WebKit/WebKit/pull/9394 1932022 18 ews-feeder 2023-02-08 08:50:25 -0800 Committed 260011@main (6ce95a8a6fc3): <https://commits.webkit.org/260011@main> Reviewed commits have been landed. Closing PR #9394 and removing active labels. 1932118 19 commit-queue 2023-02-08 13:24:34 -0800 Re-opened since this is blocked by bug 251943 1932122 20 ryanhaddad 2023-02-08 13:30:46 -0800 Reverted in https://commits.webkit.org/260032@main because it appeared to cause macOS WK1 tests to exit early with timeouts: https://build.webkit.org/results/Apple-Ventura-Release-AppleSilicon-WK1-Tests/260011@main%20(1701)/results.html 1932133 21 ryanhaddad 2023-02-08 13:56:11 -0800 This may have also caused the flakiness seen with fast/text/web-font-load-invisible-during-loading.html https://results.webkit.org/?suite=layout-tests&test=fast%2Ftext%2Fweb-font-load-invisible-during-loading.html https://build.webkit.org/results/Apple-Ventura-Release-AppleSilicon-WK2-Tests/260026@main%20(1521)/fast/text/web-font-load-invisible-during-loading-diff.txt 1932158 22 ryanhaddad 2023-02-08 15:20:08 -0800 ..and these failures on iOS Simulator: https://results.webkit.org/?suite=layout-tests&suite=layout-tests&suite=layout-tests&test=imported%2Fw3c%2Fweb-platform-tests%2Fimport-maps%2Fbare-specifiers.sub.html&test=imported%2Fw3c%2Fweb-platform-tests%2Fimport-maps%2Fdata-url-specifiers.sub.html&test=imported%2Fw3c%2Fweb-platform-tests%2Fimport-maps%2Fhttp-url-like-specifiers.sub.html&platform=ios 1932651 23 pvollan 2023-02-10 07:31:16 -0800 Pull request: https://github.com/WebKit/WebKit/pull/9925 1941633 24 ews-feeder 2023-03-15 13:02:53 -0700 Committed 261705@main (f17bb5ab82ff): <https://commits.webkit.org/261705@main> Reviewed commits have been landed. Closing PR #9925 and removing active labels. 1941634 25 435793 ap 2023-03-15 13:05:40 -0700 Comment on attachment 435793 Patch Clearing review flag, as the bug is resolved now. 434910 2021-08-04 09:41:15 -0700 2021-08-10 10:35:57 -0700 Patch bug-228780-20210804094114.patch text/plain 2167 pvollan SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDI4MDYzNSkKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDE5IEBACisyMDIxLTA4LTA0ICBQZXIgQXJu ZSBWb2xsYW4gIDxwdm9sbGFuQGFwcGxlLmNvbT4KKworICAgICAgICBUd2VhayBodG1sIHBhcnNl ciB5aWVsZGluZworICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/ aWQ9MjI4NzgwCisgICAgICAgIDxyZGFyOi8vcHJvYmxlbS84MTUxNzg0Nz4KKworICAgICAgICBS ZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBUd2VhayBodG1sIHBhcnNlciB5 aWVsZGluZyB0byBhdHRlbXB0IHRvIGltcHJvdmUgcGFnZSBsb2FkIHBlcmZvcm1hbmNlIHdpdGgg cmVzcGVjdCB0byBjZXJ0YWluIHBhZ2UgbG9hZCBtZXRyaWNzLgorICAgICAgICBDdXJyZW50bHks IHRoZSBwYXJzZXIgd2lsbCBub3QgeWllbGQgdW50aWwgMTZtcyBoYXMgZWxhcHNlZCwgd2hpY2gg Y2FuIGxlYWQgdG8gZGlmZmVyZW50IGJlaGF2aW9yIG9uIGRpZmZlcmVudAorICAgICAgICBkZXZp Y2VzLiBJbnN0ZWFkLCB0aGUgcGFyc2VyIGNhbiB5aWVsZCBpZiB0aGVyZSBjdXJyZW50bHkgaXMg YSBsYXlvdXQgcGVuZGluZyBhZnRlciB0aGUgZmlyc3QgbWVhbmluZ2Z1bCBwYWludAorICAgICAg ICBtaWxlc3RvbmUgaGFzIGJlZW4gcmVhY2hlZC4gVGhpcyB3aWxsIGxldCBsYXlvdXQgYW5kIHBh aW50aW5nIGhhdmUgcHJpb3JpdHkuCisKKyAgICAgICAgKiBodG1sL3BhcnNlci9IVE1MUGFyc2Vy U2NoZWR1bGVyLmNwcDoKKyAgICAgICAgKFdlYkNvcmU6OkhUTUxQYXJzZXJTY2hlZHVsZXI6OnNo b3VsZFlpZWxkQmVmb3JlRXhlY3V0aW5nU2NyaXB0KToKKwogMjAyMS0wOC0wNCAgQW5kcmVzIEdv bnphbGV6ICA8YW5kcmVzZ18yMkBhcHBsZS5jb20+CiAKICAgICAgICAgQWRkIHN1cHBvcnQgZm9y IGFyaWEtc2VsZWN0ZWQgdmFsdWUgY2hhbmdlcyBpbiB0YWJsZSBjZWxscy4KSW5kZXg6IFNvdXJj ZS9XZWJDb3JlL2h0bWwvcGFyc2VyL0hUTUxQYXJzZXJTY2hlZHVsZXIuY3BwCj09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0K LS0tIFNvdXJjZS9XZWJDb3JlL2h0bWwvcGFyc2VyL0hUTUxQYXJzZXJTY2hlZHVsZXIuY3BwCShy ZXZpc2lvbiAyODA0MDYpCisrKyBTb3VyY2UvV2ViQ29yZS9odG1sL3BhcnNlci9IVE1MUGFyc2Vy U2NoZWR1bGVyLmNwcAkod29ya2luZyBjb3B5KQpAQCAtMTE1LDE2ICsxMTUsMTQgQEAgYm9vbCBI VE1MUGFyc2VyU2NoZWR1bGVyOjpzaG91bGRZaWVsZEJlZgogICAgIGlmIChVTkxJS0VMWShtX2Rv Y3VtZW50SGFzQWN0aXZlUGFyc2VyWWllbGRUb2tlbnMpKQogICAgICAgICByZXR1cm4gdHJ1ZTsK IAotICAgIGF1dG8gZWxhcHNlZFRpbWUgPSBNb25vdG9uaWNUaW1lOjpub3coKSAtIHNlc3Npb24u c3RhcnRUaW1lOwotCi0gICAgY29uc3RleHByIGF1dG8gZWxhcHNlZFRpbWVMaW1pdCA9IDE2X21z OwogICAgIC8vIFJlcXVpcmUgYXQgbGVhc3Qgc29tZSBuZXcgcGFyc2VkIGNvbnRlbnQgYmVmb3Jl IHlpZWxkaW5nLgogICAgIGNvbnN0ZXhwciBhdXRvIHRva2VuTGltaXQgPSAyNTY7CiAgICAgLy8g RG9uJ3QgeWllbGQgb24gdmVyeSBzaG9ydCBpbmxpbmUgc2NyaXB0cy4gVGhpcyBpcyBhbiBpbXBl cmZlY3Qgd2F5IHRvIHRyeSB0byBndWVzcyB0aGUgZXhlY3V0aW9uIGNvc3QuCiAgICAgY29uc3Rl eHByIGF1dG8gaW5saW5lU2NyaXB0TGVuZ3RoTGltaXQgPSAxMDI0OwogCi0gICAgaWYgKGVsYXBz ZWRUaW1lIDwgZWxhcHNlZFRpbWVMaW1pdCkKKyAgICBpZiAoZG9jdW1lbnQtPnZpZXcoKSAmJiBk b2N1bWVudC0+dmlldygpLT5pc1Zpc3VhbGx5Tm9uRW1wdHkoKSAmJiAhZG9jdW1lbnQtPmlzTGF5 b3V0VGltZXJBY3RpdmUoKSkKICAgICAgICAgcmV0dXJuIGZhbHNlOworCiAgICAgaWYgKHNlc3Np b24ucHJvY2Vzc2VkVG9rZW5zIDwgdG9rZW5MaW1pdCkKICAgICAgICAgcmV0dXJuIGZhbHNlOwog Cg== 435271 2021-08-10 10:36:00 -0700 2021-08-18 13:50:14 -0700 Patch bug-228780-20210810103559.patch text/plain 2213 pvollan SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDI4MDg0OSkKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDIzIEBACisyMDIxLTA4LTEwICBQZXIgQXJu ZSAgPHB2b2xsYW5AYXBwbGUuY29tPgorCisgICAgICAgIFR3ZWFrIGh0bWwgcGFyc2VyIHlpZWxk aW5nCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0yMjg3 ODAKKyAgICAgICAgPHJkYXI6Ly9wcm9ibGVtLzgxNTE3ODQ3PgorCisgICAgICAgIFJldmlld2Vk IGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFR3ZWFrIGh0bWwgcGFyc2VyIHlpZWxkaW5n IHRvIGF0dGVtcHQgdG8gaW1wcm92ZSBwYWdlIGxvYWQgcGVyZm9ybWFuY2Ugd2l0aCByZXNwZWN0 IHRvIGNlcnRhaW4gcGFnZSBsb2FkIG1ldHJpY3MuCisgICAgICAgIFRoaXMgcGF0Y2ggbWFrZXMg dGhlIHBhcnNlciBub3QgeWllbGQgaWYgdGhlcmUgaXMgbm8gcGVuZGluZyBsYXlvdXQgaW4gb3Jk ZXIgdG8gZ2l2ZSBsYXlvdXQgYW5kIHBhaW50aW5nIGhpZ2hlcgorICAgICAgICBwcmlvcml0eS4g QWRkaXRpb25hbGx5LCB0aGlzIHlpZWxkIGNoYW5nZSBpcyBvbmx5IGFwcGxpZWQgYWZ0ZXIgYSBw YWludCBoYXMgb2NjdXJlZC4gVGhlIHJlYXNvbiBmb3IgdGhpcyBpcyB0aGF0CisgICAgICAgIHBh cnNlciB5aWVsZGluZyBhbHJlYWR5IHNlZW1zIHRvIGJlIGhpZ2hseSBvcHRpbXplZCB3aXRoIHJl Z2FyZHMgdG8gZmlyc3QgcGFpbnQgYW5kIGZpcnN0IG1lYW5pbmdmdWwgcGFpbnQsIHNvCisgICAg ICAgIGFueSB5aWVsZCBjaGFuZ2UgaW4gdGhpcyB0aW1lIGludGVydmFsIHdvdWxkIGxpa2VseSBi ZSBhIHBlcmZvcm1hbmNlIHJlZ3Jlc3Npb24uIEl0IGlzIG9ubHkgYWZ0ZXIgdGhlc2UgZXZlbnRz CisgICAgICAgIHRoZXJlIHNlZW1zIHRvIGJlIGEgc2xpZ2h0IHJvb20gZm9yIHBlcmZvcm1hbmNl IGltcHJvdmVtZW50LgorCisgICAgICAgIE5vIG5ldyB0ZXN0cywgY292ZXJlZCBieSBleGl0aW5n IHRlc3RzLgorCisgICAgICAgICogaHRtbC9wYXJzZXIvSFRNTFBhcnNlclNjaGVkdWxlci5jcHA6 CisgICAgICAgIChXZWJDb3JlOjpIVE1MUGFyc2VyU2NoZWR1bGVyOjpzaG91bGRZaWVsZEJlZm9y ZUV4ZWN1dGluZ1NjcmlwdCk6CisKIDIwMjEtMDgtMTAgIERldmluIFJvdXNzbyAgPGRyb3Vzc29A YXBwbGUuY29tPgogCiAgICAgICAgIFttYWNPU10gUkVHUkVTU0lPTihyMjc4ODUwKTogbW9kaWZ5 aW5nIGBwbGF5YmFja1JhdGVgIHZpYSBKUyBvbiBhIFRvdWNoQmFyIG1hYyBvbmx5IHNvbWV0aW1l cyB3b3JrcyBhbmQgYWxzbyBjYXVzZXMgdGhlIGBkZWZhdWx0UGxheWJhY2tSYXRlYCB0byBjaGFu Z2UKSW5kZXg6IFNvdXJjZS9XZWJDb3JlL2h0bWwvcGFyc2VyL0hUTUxQYXJzZXJTY2hlZHVsZXIu Y3BwCj09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT0KLS0tIFNvdXJjZS9XZWJDb3JlL2h0bWwvcGFyc2VyL0hUTUxQYXJzZXJT Y2hlZHVsZXIuY3BwCShyZXZpc2lvbiAyODA3MzQpCisrKyBTb3VyY2UvV2ViQ29yZS9odG1sL3Bh cnNlci9IVE1MUGFyc2VyU2NoZWR1bGVyLmNwcAkod29ya2luZyBjb3B5KQpAQCAtMTIzLDYgKzEy Myw5IEBAIGJvb2wgSFRNTFBhcnNlclNjaGVkdWxlcjo6c2hvdWxkWWllbGRCZWYKICAgICAvLyBE b24ndCB5aWVsZCBvbiB2ZXJ5IHNob3J0IGlubGluZSBzY3JpcHRzLiBUaGlzIGlzIGFuIGltcGVy ZmVjdCB3YXkgdG8gdHJ5IHRvIGd1ZXNzIHRoZSBleGVjdXRpb24gY29zdC4KICAgICBjb25zdGV4 cHIgYXV0byBpbmxpbmVTY3JpcHRMZW5ndGhMaW1pdCA9IDEwMjQ7CiAKKyAgICBpZiAoZG9jdW1l bnQtPnZpZXcoKSAmJiBkb2N1bWVudC0+dmlldygpLT5oYXNFdmVyUGFpbnRlZCgpICYmICFkb2N1 bWVudC0+aXNMYXlvdXRQZW5kaW5nKCkpCisgICAgICAgIHJldHVybiBmYWxzZTsKKwogICAgIGlm IChlbGFwc2VkVGltZSA8IGVsYXBzZWRUaW1lTGltaXQpCiAgICAgICAgIHJldHVybiBmYWxzZTsK ICAgICBpZiAoc2Vzc2lvbi5wcm9jZXNzZWRUb2tlbnMgPCB0b2tlbkxpbWl0KQo= 435793 2021-08-18 13:50:17 -0700 2023-03-15 13:05:40 -0700 Patch bug-228780-20210818135016.patch text/plain 2252 pvollan SW5kZXg6IFNvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2Vi Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDI4MDc4NCkKKysrIFNvdXJjZS9XZWJDb3JlL0NoYW5n ZUxvZwkod29ya2luZyBjb3B5KQpAQCAtMSwzICsxLDI1IEBACisyMDIxLTA4LTE4ICBQZXIgQXJu ZSAgPHB2b2xsYW5AYXBwbGUuY29tPgorCisgICAgICAgIFR3ZWFrIGh0bWwgcGFyc2VyIHlpZWxk aW5nCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0yMjg3 ODAKKyAgICAgICAgPHJkYXI6Ly9wcm9ibGVtLzgxNTE3ODQ3PgorCisgICAgICAgIFJldmlld2Vk IGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAgIFR3ZWFrIGh0bWwgcGFyc2VyIHlpZWxkaW5n IHRvIGF0dGVtcHQgdG8gaW1wcm92ZSBwYWdlIGxvYWQgcGVyZm9ybWFuY2Ugd2l0aCByZXNwZWN0 IHRvIGNlcnRhaW4gcGFnZSBsb2FkIG1ldHJpY3MuCisgICAgICAgIFRoaXMgcGF0Y2ggbWFrZXMg dGhlIHBhcnNlciBub3QgeWllbGQgaWYgdGhlcmUgaXMgbm8gcGVuZGluZyBsYXlvdXQgaW4gb3Jk ZXIgdG8gZ2l2ZSBsYXlvdXQgYW5kIHBhaW50aW5nIGhpZ2hlcgorICAgICAgICBwcmlvcml0eS4g VGhpcyB5aWVsZCBjaGFuZ2UgaXMgb25seSBhcHBsaWVkIGFmdGVyIGEgcGFpbnQgaGFzIG9jY3Vy ZWQuIFRoZSByZWFzb24gZm9yIHRoaXMgaXMgdGhhdCBwYXJzZXIgeWllbGRpbmcKKyAgICAgICAg YWxyZWFkeSBzZWVtcyB0byBiZSBoaWdobHkgb3B0aW1pemVkIHdpdGggcmVnYXJkcyB0byBmaXJz dCBwYWludCBhbmQgZmlyc3QgbWVhbmluZ2Z1bCBwYWludCwgc28gYW55IHlpZWxkIGNoYW5nZQor ICAgICAgICBpbiB0aGlzIHRpbWUgaW50ZXJ2YWwgd291bGQgbGlrZWx5IGJlIGEgcGVyZm9ybWFu Y2UgcmVncmVzc2lvbi4gSXQgaXMgb25seSBhZnRlciB0aGVzZSBldmVudHMgdGhlcmUgc2VlbXMg dG8gYmUgYQorICAgICAgICBzbGlnaHQgcm9vbSBmb3IgcGVyZm9ybWFuY2UgaW1wcm92ZW1lbnQu IEZpbmFsbHksIGEgdGhpcmQgY29uZGl0aW9uIGlzIGFkZGVkOyBuYW1lbHkgdGhhdCB5aWVsZGlu ZyBzaG91bGQgb25seSBiZQorICAgICAgICBhdm9pZGVkIGZvciBpbmxpbmUgc2NyaXB0cy4gVGhp cyBsYXN0IG1vZGlmaWNhdGlvbiB3YXMgZG9uZSBpbiBvcmRlciB0byBhdm9pZCByZWdyZXNzaW9u cyBpbiBvdGhlciBwYWdlIGxvYWRpbmcKKyAgICAgICAgbWV0cmljcyB3ZSBhcmUgbWVhc3VyaW5n LgorCisgICAgICAgIE5vIG5ldyB0ZXN0cywgY292ZXJlZCBieSBleGl0aW5nIHRlc3RzLgorCisg ICAgICAgICogaHRtbC9wYXJzZXIvSFRNTFBhcnNlclNjaGVkdWxlci5jcHA6CisgICAgICAgIChX ZWJDb3JlOjpIVE1MUGFyc2VyU2NoZWR1bGVyOjpzaG91bGRZaWVsZEJlZm9yZUV4ZWN1dGluZ1Nj cmlwdCk6CisKIDIwMjEtMDgtMDkgIERlYW4gSmFja3NvbiAgPGRpbm9AYXBwbGUuY29tPgogCiAg ICAgICAgIFdlYkdMIDIuMCBkb2Vzbid0IHdvcmsgb24gQTggZGV2aWNlcyAoaVBhZCBtaW5pIDQs IGlQYWQgQWlyIDIpCkluZGV4OiBTb3VyY2UvV2ViQ29yZS9odG1sL3BhcnNlci9IVE1MUGFyc2Vy U2NoZWR1bGVyLmNwcAo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBTb3VyY2UvV2ViQ29yZS9odG1sL3BhcnNlci9I VE1MUGFyc2VyU2NoZWR1bGVyLmNwcAkocmV2aXNpb24gMjgwNzg0KQorKysgU291cmNlL1dlYkNv cmUvaHRtbC9wYXJzZXIvSFRNTFBhcnNlclNjaGVkdWxlci5jcHAJKHdvcmtpbmcgY29weSkKQEAg LTEzNiw2ICsxMzYsMTEgQEAgYm9vbCBIVE1MUGFyc2VyU2NoZWR1bGVyOjpzaG91bGRZaWVsZEJl ZgogICAgICAgICAgICAgcmV0dXJuIGZhbHNlOwogICAgIH0KIAorICAgIGJvb2wgaW5saW5lU2Ny aXB0ID0gc2NyaXB0RWxlbWVudCAmJiAhc2NyaXB0RWxlbWVudC0+aGFzU291cmNlQXR0cmlidXRl KCk7CisgICAgYm9vbCBoYXNQYWludGVkID0gZG9jdW1lbnQtPnZpZXcoKSAmJiBkb2N1bWVudC0+ dmlldygpLT5oYXNFdmVyUGFpbnRlZCgpOworICAgIGlmIChoYXNQYWludGVkICYmICFkb2N1bWVu dC0+aXNMYXlvdXRQZW5kaW5nKCkgJiYgaW5saW5lU2NyaXB0KQorICAgICAgICByZXR1cm4gZmFs c2U7CisKICAgICByZXR1cm4gdHJ1ZTsKIH0KIAo=