226330 2021-05-27 06:55:04 -0700 Double clicking on very large but visually empty DOM can be slow due to selection computation 2022-06-28 01:47:24 -0700 1 1 1 Unclassified WebKit CSS Safari Technology Preview Unspecified Unspecified NEW InRadar P2 Normal --- 1 ldebeasi webkit-unassigned emilio koivisto marc phil.moore rniwa simon.fraser webkit-bug-importer zalan oldest_to_newest 1764316 0 429874 ldebeasi 2021-05-27 06:55:04 -0700 Created attachment 429874 Code reproduction This is a continuation of https://bugs.webkit.org/show_bug.cgi?id=222187 which was partially resolved in STP 125. When using Web Components inside of the Shadow DOM that have CSS Variables set on the host, there is a significant delay in response to clicks/touches. There were fixes in STP 125 that improved this issue, but I am still able to reproduce this issue in other ways. Steps to reproduce: 1. Open code reproduction in STP 125. 2. Quickly click multiple times on the "Click Me" label (not the checkbox) and you should see Safari freeze up. 3. If you repeat step 2 while taking a Timeline sample, you should see the CPU usage spike significantly. In my tests I clicked about 4-5 times and saw the CPU spike to over 80% on average. Expected Behavior: I would expect Safari to not freeze up when quickly clicking the label multiple times. Actual Behavior: Safari freezes up when quickly clicking the label multiple times. Other Info: - As noted in https://bugs.webkit.org/show_bug.cgi?id=222187, the example I used is inherently heavy in shadow trees to emphasize the issue. Please see https://github.com/ionic-team/ionic-framework/issues/22951 for more "real world" examples of this impacting applications. 1764619 1 ap 2021-05-27 18:21:22 -0700 Thank you for the report! Is the performance still worse than pre-iOS 14, or is this about further optimization opportunities? Does this appear to be worse in Safari (STP) than in other browsers? 1764847 2 ldebeasi 2021-05-28 11:37:16 -0700 The performance of single clicking the checkbox in my example is good after https://bugs.webkit.org/show_bug.cgi?id=222187, but multiple clicks is where there are still issues. When doing multiple clicks, the performance in STP 125 is worse than in other browsers and worse than in pre-iOS/Safari 14. 1766374 3 webkit-bug-importer 2021-06-03 06:56:16 -0700 <rdar://problem/78813010> 1772626 4 marc 2021-06-25 05:30:12 -0700 Hi guys, just wondered if there's been any movement on this? Thanks, 1876997 5 phil.moore 2022-06-21 10:34:25 -0700 Be great to understand when this might get resolved, we’ve got a production app https://apps.apple.com/gb/app/mydrivetime-companion-app/id950711705 that’s being heavily impacted and stopping us from releasing a major update that our users need. Many thanks 1877248 6 marc 2022-06-22 01:10:26 -0700 (In reply to Philip Moore from comment #5) > Be great to understand when this might get resolved, we’ve got a production > app https://apps.apple.com/gb/app/mydrivetime-companion-app/id950711705 > that’s being heavily impacted and stopping us from releasing a major update > that our users need. > > Many thanks Hi Philip, by the fact this issue hasn't really moved forwards in over a year, I don't think anyone at webkit really cares about it. We had exactly the same issue last year and they've essentially ignored this thread. It seems we are expected to have write overly exuberant lazy loading component code to over come an issue that shouldn't be a problem to start with, as any other browser/device works fine, it's just anything Apple. This is ridiculous, if I have an issue open for more than a 2 weeks my bosses complain, I think I would have been fired if I had a bug open for over a year. 1877293 7 ldebeasi 2022-06-22 08:22:11 -0700 Hey everyone, While I know it is frustrating to not have your ticket resolved quickly, comments like the one in https://bugs.webkit.org/show_bug.cgi?id=226330#c6 are not super constructive. Instead, comments that provide context around the impact of a bug (such as the comment in https://bugs.webkit.org/show_bug.cgi?id=226330#c5) are incredibly helpful. This context can be used to better prioritize issues. Having worked on a large open source project (though not as large as a browser engine), I can say that it is not always possible to reply to every single issue/comment. Let’s try to keep this thread constructive and focus on identifying the impact this bug has on production apps/websites. 1877554 8 zalan 2022-06-22 21:43:21 -0700 I did a bit of profiling on this and it looks like (after the invalidation fix), this is simply about iterating over that massive amount of content when computing selection boundary. The following, simple markup triggers this issue: <body> Click Me<input type="checkbox"> repeat it for a few thousands of times </body> It looks like VisibleSelection needs a more efficient way to figure out the selection end on user action (or in general for that matter). 1877657 9 koivisto 2022-06-23 09:03:13 -0700 Note that you can work around this issue by preventing the selection with "user-select:none". 1877677 10 zalan 2022-06-23 10:03:31 -0700 (In reply to zalan from comment #8) > I did a bit of profiling on this and it looks like (after the invalidation > fix), this is simply about iterating over that massive amount of content > when computing selection boundary. > The following, simple markup triggers this issue: > > <body> > Click Me<input type="checkbox"> > > > > > > repeat it for a few thousands of times > </body> > > It looks like VisibleSelection needs a more efficient way to figure out the > selection end on user action (or in general for that matter). However if there's any "non-visually empty" content somewhere e.g. <body> Click Me<input type="checkbox"> !!!some text!!! repeat it for a few thousand times </body> I don't see any spinning anymore. So this is about iterating over massive amount of _visually empty_ content (see nextVisuallyDistinctCandidate()). 1878595 11 rniwa 2022-06-27 14:41:30 -0700 Sounds like this is unrelated to shadow DOM. It's a good bug to address nonetheless. 429874 2021-05-27 06:55:04 -0700 2021-05-27 06:55:04 -0700 Code reproduction index.html text/html 2846 ldebeasi CjwhRE9DVFlQRSBodG1sPgo8aHRtbD4KICA8aGVhZD4KICAgIDx0aXRsZT5TaGFkb3cgVGVzdDwv dGl0bGU+CiAgICA8bWV0YSBuYW1lPSJ2aWV3cG9ydCIgY29udGVudD0id2lkdGg9ZGV2aWNlLXdp ZHRoLCBpbml0aWFsLXNjYWxlPTEuMCwgbWluaW11bS1zY2FsZT0xLjAsIG1heGltdW0tc2NhbGU9 MS4wLCB1c2VyLXNjYWxhYmxlPW5vIj4KICA8L2hlYWQ+Cgo8Ym9keT4KICA8YXBwLWNvbnRlbnQg Y2xhc3M9ImNvbnRlbnQiPjwvYXBwLWNvbnRlbnQ+CgogIDxzY3JpcHQ+CiAgY2xhc3MgQXBwQ29u dGVudCBleHRlbmRzIEhUTUxFbGVtZW50IHsKICAgIGNvbnN0cnVjdG9yKCkgewogICAgICBzdXBl cigpOwogICAgICAKICAgICAgdGhpcwogICAgICAgIC5hdHRhY2hTaGFkb3coeyBtb2RlOiAnb3Bl bicgfSkKICAgICAgICAuaW5uZXJIVE1MID0gYDxzbG90Pjwvc2xvdD5gOyAgIAogICAgfQogIH0K ICAKICBjbGFzcyBBcHBJdGVtIGV4dGVuZHMgSFRNTEVsZW1lbnQgewogICAgY29uc3RydWN0b3Io KSB7CiAgICAgIHN1cGVyKCk7CgogICAgICB0aGlzCiAgICAgICAgLmF0dGFjaFNoYWRvdyh7IG1v ZGU6ICdvcGVuJyB9KQogICAgICAgIC5pbm5lckhUTUwgPSBgCiAgICAgICAgPHN0eWxlPgogICAg ICAgIDpob3N0IHsKICAgICAgICAgIC0tbXktdmFyLWE6IDBweDsKICAgICAgICAgIC0tbXktdmFy LWI6IDBweDsKICAgICAgICAgIC0tbXktdmFyLWM6IDBweDsKICAgICAgICAgIC0tbXktdmFyLWQ6 IDBweDsKICAgICAgICAgIC0tbXktdmFyLWU6IDBweDsKICAgICAgICAgIC0tbXktdmFyLWY6IDBw eDsKICAgICAgICAgIC0tbXktdmFyLWc6IDBweDsKICAgICAgICAgIC0tbXktdmFyLWg6IDBweDsK ICAgICAgICAgIC0tbXktdmFyLWk6IDBweDsKICAgICAgICAgIC0tbXktdmFyLWo6IDBweDsKICAg ICAgICAgIC0tbXktdmFyLWs6IDBweDsKICAgICAgICAgIC0tbXktdmFyLWw6IDBweDsKICAgICAg ICAgIC0tbXktdmFyLW06IDBweDsKICAgICAgICAgIC0tbXktdmFyLW46IDBweDsKICAgICAgICAg IC0tbXktdmFyLW86IDBweDsKICAgICAgICAgIC0tbXktdmFyLXA6IDBweDsKICAgICAgICAgIC0t bXktdmFyLXE6IDBweDsKICAgICAgICAgIC0tbXktdmFyLXI6IDBweDsKICAgICAgICAgIC0tbXkt dmFyLXM6IDBweDsKICAgICAgICAgIC0tbXktdmFyLXQ6IDBweDsKICAgICAgICAgIC0tbXktdmFy LXU6IDBweDsKICAgICAgICAgIC0tbXktdmFyLXY6IDBweDsKICAgICAgICAgIC0tbXktdmFyLXc6 IDBweDsKICAgICAgICAgIC0tbXktdmFyLXg6IDBweDsKICAgICAgICAgIC0tbXktdmFyLXk6IDBw eDsKICAgICAgICAgIC0tbXktdmFyLXo6IDBweDsKICAgICAgICB9CiAgICAgICAgPC9zdHlsZT4K ICAgICAgICA8c2xvdD48L3Nsb3Q+YDsgICAKICAgIH0KICB9CgogIAogIGN1c3RvbUVsZW1lbnRz LmRlZmluZSgnYXBwLWNvbnRlbnQnLCBBcHBDb250ZW50KTsKICBjdXN0b21FbGVtZW50cy5kZWZp bmUoJ2FwcC1pdGVtJywgQXBwSXRlbSk7CiAgCiAgY29uc3QgY29udGVudCA9IGRvY3VtZW50LnF1 ZXJ5U2VsZWN0b3IoJy5jb250ZW50Jyk7CiAgbGV0IHRlbXBsYXRlID0gYAogIDxsYWJlbCBmb3I9 Im15LWNoZWNrYm94Ij5DbGljayBNZTwvbGFiZWw+CiAgPGlucHV0IHR5cGU9ImNoZWNrYm94IiBp ZD0ibXktY2hlY2tib3giIC8+IDxiciAvPgogIGA7CgogIGZvciAobGV0IGkgPSAwOyBpIDwgMTUw OyBpKyspIHsKICAgIHRlbXBsYXRlICs9IGAKICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgogICAg PGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAt aXRlbT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+ PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBw LWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVt PgogICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAg IDxhcHAtaXRlbT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBw LWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVt PjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2Fw cC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRl bT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAg ICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgogICAgPGFw cC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRl bT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9h cHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBwLWl0 ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgog ICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0ZW0+PC9hcHAtaXRlbT4KICAgIDxh cHAtaXRlbT48L2FwcC1pdGVtPgogICAgPGFwcC1pdGVtPjwvYXBwLWl0ZW0+CiAgICA8YXBwLWl0 ZW0+PC9hcHAtaXRlbT4KICAgIDxhcHAtaXRlbT48L2FwcC1pdGVtPgpgOwogIH0KCiAgY29udGVu dC5pbm5lckhUTUwgPSB0ZW1wbGF0ZTsKICA8L3NjcmlwdD4KCjwvYm9keT4KPC9odG1sPgo=