305256 2026-01-10 03:03:21 -0800 LCP buffer size seems to be set to 1 instead of 150 2026-04-30 14:46:16 -0700 1 1 1 Unclassified WebKit Layout and Rendering Safari 26 Mac (Apple Silicon) macOS 15 RESOLVED FIXED InRadar P2 Normal --- 1 barry simon.fraser bfulgham karlcow ryan simon.fraser webkit-bug-importer zalan oldest_to_newest 2170893 0 barry 2026-01-10 03:03:21 -0800 This page loads some text and then a delayed `<img>` https://www.tunetheweb.com/experiments/lcp/h1-image.html The LCP is correctly reports LCP candidates as: ``` h1#heading img ``` These are got from and LCP performance observer - basically this code: ```js new PerformanceObserver((entryList) => { for (const entry of entryList.getEntries()) { console.log('LCP candidate:', entry.startTime, entry); } }).observe({type: 'largest-contentful-paint', buffered: true}); ``` This page also loads some text and then a delayed `<img>`, but delays starting the performance observer for two seconds (after the image has loaded): https://www.tunetheweb.com/experiments/lcp/h1-image.html In this case the LCP candidates only contains the second image: ``` img ``` While the largest element is usually the most important for LCP, it's useful for debugging purposes to understand if LCP changed over time, even if a performance observer was setup intitall. The registry says LCP should have a buffer size of 150: https://w3c.github.io/timing-entrytypes-registry/#registry 2170894 1 barry 2026-01-10 03:04:38 -0800 Urgh, sorry that second test URL should be: https://www.tunetheweb.com/experiments/lcp/h1-image-delay.html 2171262 2 webkit-bug-importer 2026-01-12 13:51:53 -0800 <rdar://problem/168015751> 2171324 3 simon.fraser 2026-01-12 19:56:00 -0800 Sure would be nice if https://www.w3.org/TR/largest-contentful-paint/ referenced https://w3c.github.io/timing-entrytypes-registry/#registry 2171351 4 barry 2026-01-12 23:32:45 -0800 Agreed! Raised this: https://github.com/w3c/largest-contentful-paint/issues/157 2173492 5 simon.fraser 2026-01-20 16:10:42 -0800 https://www.tunetheweb.com/experiments/lcp/h1-image.html gives me ``` h1#heading img ``` in Safari, and STP. Also, in a standalone test that tries to exercise the buffer limit, Chrome seems to have a buffer size of 1. Do you have a test that explicitly tests the buffer size? 2173498 6 478082 simon.fraser 2026-01-20 16:26:37 -0800 Created attachment 478082 Prospective test Adding a WPT that I think should show that the buffer size is > 1. 2173528 7 barry 2026-01-20 17:38:10 -0800 > Do you have a test that explicitly tests the buffer size? Yes as per comment https://bugs.webkit.org/show_bug.cgi?id=305256#c1 the following test gives different results in Chrome and Safari: https://www.tunetheweb.com/experiments/lcp/h1-image-delay.html 2173535 8 simon.fraser 2026-01-20 18:07:40 -0800 Got it, sorry I missed that (oh for editable comments) 2183167 9 simon.fraser 2026-02-19 15:04:40 -0800 We need an implementation of https://w3c.github.io/performance-timeline/#dfn-queue-a-performanceentry shared between different performance entry types. 2205617 10 simon.fraser 2026-04-28 18:47:32 -0700 Pull request: https://github.com/WebKit/WebKit/pull/63862 2206391 11 ews-feeder 2026-04-30 14:46:14 -0700 Committed 312393@main (86af78950458): <https://commits.webkit.org/312393@main> Reviewed commits have been landed. Closing PR #63862 and removing active labels. 478082 2026-01-20 16:26:37 -0800 2026-01-20 16:26:37 -0800 Prospective test lcp-buffer-size.html text/html 1637 simon.fraser PCFET0NUWVBFIEhUTUw+CjxtZXRhIGNoYXJzZXQ9dXRmLTg+Cjx0aXRsZT5MYXJnZXN0Q29udGVu dGZ1bFBhaW50IGNvbXBhcmVkIHdpdGggRmlyc3RQYWludCBhbmQgRmlyc3RDb250ZW50ZnVsUGFp bnQgb24gc2luZ2xlIHRleHQgcGFnZS48L3RpdGxlPgo8c3R5bGU+CiAgI2NvbnRhaW5lciA+IGRp diB7CiAgICBtYXJnaW46IDJweDsKICAgIGJhY2tncm91bmQtY29sb3I6IGJsdWU7CiAgfQo8L3N0 eWxlPgo8Ym9keT4KPHNjcmlwdCBzcmM9Ii9yZXNvdXJjZXMvdGVzdGhhcm5lc3MuanMiPjwvc2Ny aXB0Pgo8c2NyaXB0IHNyYz0iL3Jlc291cmNlcy90ZXN0aGFybmVzc3JlcG9ydC5qcyI+PC9zY3Jp cHQ+CjxzZWN0aW9uIGlkPSJjb250YWluZXIiPjwvc2VjdGlvbj4KPHNjcmlwdD4KICBwcm9taXNl X3Rlc3QoYXN5bmMgKHQpID0+IHsKICAgIGFzc2VydF9pbXBsZW1lbnRzKHdpbmRvdy5QZXJmb3Jt YW5jZVBhaW50VGltaW5nLCAiUGVyZm9ybWFuY2VQYWludFRpbWluZyBpcyBub3QgaW1wbGVtZW50 ZWQiKTsKICAgIGFzc2VydF9pbXBsZW1lbnRzKHdpbmRvdy5MYXJnZXN0Q29udGVudGZ1bFBhaW50 LCAiTGFyZ2VzdENvbnRlbnRmdWxQYWludCBpcyBub3QgaW1wbGVtZW50ZWQiKTsKICAgIAogICAg Y29uc3QgbWF4TENQQ291bnQgPSAxMDsKICAgIAogICAgbGV0IGN1cnJTaXplID0gNDsKICAgIGNv bnN0IGNvbnRhaW5lciA9IGRvY3VtZW50LmdldEVsZW1lbnRCeUlkKCdjb250YWluZXInKTsKICAg IAogICAgZm9yIChsZXQgaSA9IDA7IGkgPCBtYXhMQ1BDb3VudDsgKytpKSB7CiAgICAgIGNvbnN0 IGRpdiA9IGRvY3VtZW50LmNyZWF0ZUVsZW1lbnQoJ2RpdicpOwogICAgICBkaXYuaWQgPSBgaXRl bV8ke2N1cnJTaXplfWA7CiAgICAgIGRpdi5zdHlsZSA9IGBiYWNrZ3JvdW5kLWltYWdlOgogICAg ICAgIHVybCgvaW1hZ2VzL2xjcC05Nng5Ni5wbmc/JHtjdXJyU2l6ZX0pOyB3aWR0aDogJHtjdXJy U2l6ZX1weDsgaGVpZ2h0OiAke2N1cnJTaXplfXB4YDsKCiAgICAgICsrY3VyclNpemU7CiAgICAg IAogICAgICBjb250YWluZXIuYXBwZW5kQ2hpbGQoZGl2KTsKCiAgICAgIGF3YWl0IG5ldyBQcm9t aXNlKHJlcXVlc3RBbmltYXRpb25GcmFtZSk7CiAgICB9CgogICAgbGV0IGVudHJ5TGlzdExlbmd0 aCA9IGF3YWl0IG5ldyBQcm9taXNlKHJlc29sdmUgPT4gewogICAgICBuZXcgUGVyZm9ybWFuY2VP YnNlcnZlcigoZW50cnlMaXN0LCBvYnNlcnZlcikgPT4gewoKICAgICAgICBpZiAoZW50cnlMaXN0 LmdldEVudHJpZXMoKS5maWx0ZXIoZSA9PiBlLmlkICE9PSAnJykubGVuZ3RoID4gMCkgewogICAg ICAgICAgcmVzb2x2ZShlbnRyeUxpc3QuZ2V0RW50cmllcygpLmxlbmd0aCk7CiAgICAgICAgICBv YnNlcnZlci5kaXNjb25uZWN0KCk7CiAgICAgICAgfQoKICAgICAgfSkub2JzZXJ2ZSh7IHR5cGU6 ICdsYXJnZXN0LWNvbnRlbnRmdWwtcGFpbnQnLCBidWZmZXJlZDogdHJ1ZSB9KTsKICAgIH0pOwoK ICAgIGFzc2VydF9lcXVhbHMoZW50cnlMaXN0TGVuZ3RoLCBtYXhMQ1BDb3VudCwgJ1RoZXJlIHNo b3VsZCBiZSBvbmUgYW5kIG9ubHkgb25lIExDUCBlbnRyeS4nKTsKCiAgfSwgJ0xDUCBidWZmZXIg c2l6ZSBpcyBhdCBsZWFzdCAxMCcpOwo8L3NjcmlwdD4KPC9ib2R5Pgo=