105512 2012-12-20 01:47:34 -0800 Avoid a second layout of flex items in computeMainAxisPreferredSizes 2017-07-18 08:29:30 -0700 1 1 1 Unclassified WebKit Layout and Rendering 528+ (Nightly build) Unspecified Unspecified NEW P2 Normal --- 62048 1 cgarcia webkit-unassigned eric ojan.autocc tony webkit.review.bot oldest_to_newest 795282 0 cgarcia 2012-12-20 01:47:34 -0800 In case of orthogonal flow when the flex basis or min main axis are auto for child, we are laying out the child in computeMainAxisPreferredSizes. If called a second time, we don't need to re-layout the child unless its main axis size is a percentage. 795286 1 180302 cgarcia 2012-12-20 01:49:40 -0800 Created attachment 180302 Patch 795532 2 tony 2012-12-20 10:09:36 -0800 It seems like if the flex size is different from the preferred size, during a relayout, you'll use the previous flex size instead of recomputing the preferred size. I'm not sure we have a test case that does this, but we should try to add one. Another idea would be to cache the child's preferred size and if we don't need to relayout the child, we could use the cached preferred size. Not sure how much of a win that really is. 795632 3 tony 2012-12-20 11:58:47 -0800 Here's an example: http://plexode.com/eval3/#ht=%3Cdiv%20id%3D%22container%22%20style%3D%22height%3A%20200px%22%3E%0A%3Cdiv%20style%3D%22display%3A-webkit-flex%3B%20-webkit-flex-flow%3A%20column%3B%20height%3A%20100%25%22%3E%0A%20%20%3Cdiv%20id%3D%22a%22%20style%3D%22-webkit-flex%3A%201%20auto%3B%20background-color%3A%20pink%3B%22%3Ea%3C%2Fdiv%3E%0A%20%20%3Cdiv%20id%3Db%20style%3D%22-webkit-flex%3A%201%2050px%3B%20background-color%3A%20lime%3B%22%3Eb%3C%2Fdiv%3E%0A%3C%2Fdiv%3E%0A%3C%2Fdiv%3E&ohj=0&jt=container.style.height%20%3D%20%22100px%22%0AgetComputedStyle(a).height%20%2B%20%22%2C%22%20%2B%20getComputedStyle(b).height&ojh=0&ojj=0 Press the "eval JS now" button. It reduces the height of the container, but we don't relayout the flex children, so they stay the same size. 796128 4 180302 buildbot 2012-12-20 21:29:42 -0800 Comment on attachment 180302 Patch Attachment 180302 did not pass mac-ews (mac): Output: http://queues.webkit.org/results/15460195 New failing tests: css3/flexbox/flexitem.html 796270 5 180302 cgarcia 2012-12-21 06:14:52 -0800 Comment on attachment 180302 Patch Clearing flags since the patch is not correct 180302 2012-12-20 01:49:40 -0800 2012-12-21 06:14:52 -0800 Patch wcore-flexbox-relayout.diff text/plain 1988 cgarcia ZGlmZiAtLWdpdCBhL1NvdXJjZS9XZWJDb3JlL0NoYW5nZUxvZyBiL1NvdXJjZS9XZWJDb3JlL0No YW5nZUxvZwppbmRleCA2YTUxOTZmLi44YzFiNWRiIDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViQ29y ZS9DaGFuZ2VMb2cKKysrIGIvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTkg QEAKKzIwMTItMTItMjAgIENhcmxvcyBHYXJjaWEgQ2FtcG9zICA8Y2dhcmNpYUBpZ2FsaWEuY29t PgorCisgICAgICAgIEF2b2lkIGEgc2Vjb25kIGxheW91dCBvZiBmbGV4IGl0ZW1zIGluIGNvbXB1 dGVNYWluQXhpc1ByZWZlcnJlZFNpemVzCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3Jn L3Nob3dfYnVnLmNnaT9pZD0xMDU1MTIKKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9P UFMhKS4KKworICAgICAgICBJbiBjYXNlIG9mIG9ydGhvZ29uYWwgZmxvdyB3aGVuIHRoZSBmbGV4 IGJhc2lzIG9yIG1pbiBtYWluIGF4aXMKKyAgICAgICAgYXJlIGF1dG8gZm9yIGNoaWxkLCB3ZSBh cmUgbGF5aW5nIG91dCB0aGUgY2hpbGQgaW4KKyAgICAgICAgY29tcHV0ZU1haW5BeGlzUHJlZmVy cmVkU2l6ZXMuIElmIGNhbGxlZCBhIHNlY29uZCB0aW1lLCB3ZSBkb24ndAorICAgICAgICBuZWVk IHRvIHJlLWxheW91dCB0aGUgY2hpbGQgdW5sZXNzIGl0cyBtYWluIGF4aXMgc2l6ZSBpcyBhCisg ICAgICAgIHBlcmNlbnRhZ2UuCisKKyAgICAgICAgKiByZW5kZXJpbmcvUmVuZGVyRmxleGlibGVC b3guY3BwOgorICAgICAgICAoV2ViQ29yZTo6UmVuZGVyRmxleGlibGVCb3g6OmNvbXB1dGVNYWlu QXhpc1ByZWZlcnJlZFNpemVzKToKKwogMjAxMi0xMi0xOSAgQ2FybG9zIEdhcmNpYSBDYW1wb3Mg IDxjZ2FyY2lhQGlnYWxpYS5jb20+CiAKICAgICAgICAgTWFrZSBvcmRlciBpdGVyYXRvciBtZW1i ZXIgc3RhY2sgYWxsb2NhdGVkIGluIFJlbmRlckZsZXhpYmxlQm94CmRpZmYgLS1naXQgYS9Tb3Vy Y2UvV2ViQ29yZS9yZW5kZXJpbmcvUmVuZGVyRmxleGlibGVCb3guY3BwIGIvU291cmNlL1dlYkNv cmUvcmVuZGVyaW5nL1JlbmRlckZsZXhpYmxlQm94LmNwcAppbmRleCBjZGIyYTExLi44MTc0MjU1 IDEwMDY0NAotLS0gYS9Tb3VyY2UvV2ViQ29yZS9yZW5kZXJpbmcvUmVuZGVyRmxleGlibGVCb3gu Y3BwCisrKyBiL1NvdXJjZS9XZWJDb3JlL3JlbmRlcmluZy9SZW5kZXJGbGV4aWJsZUJveC5jcHAK QEAgLTg2MSw3ICs4NjEsOSBAQCB2b2lkIFJlbmRlckZsZXhpYmxlQm94Ojpjb21wdXRlTWFpbkF4 aXNQcmVmZXJyZWRTaXplcyhPcmRlckhhc2hTZXQmIG9yZGVyVmFsdWVzKQogICAgICAgICAvLyBP bmx5IG5lZWQgdG8gbGF5b3V0IGhlcmUgaWYgd2Ugd2lsbCBuZWVkIHRvIGdldCB0aGUgbG9naWNh bEhlaWdodCBvZiB0aGUgY2hpbGQgaW4gY29tcHV0ZU5leHRGbGV4TGluZS4KICAgICAgICAgTGVu Z3RoIGNoaWxkTWFpbkF4aXNNaW4gPSBpc0hvcml6b250YWxGbG93KCkgPyBjaGlsZC0+c3R5bGUo KS0+bWluV2lkdGgoKSA6IGNoaWxkLT5zdHlsZSgpLT5taW5IZWlnaHQoKTsKICAgICAgICAgaWYg KGhhc09ydGhvZ29uYWxGbG93KGNoaWxkKSAmJiAoZmxleEJhc2lzRm9yQ2hpbGQoY2hpbGQpLmlz QXV0bygpIHx8IGNoaWxkTWFpbkF4aXNNaW4uaXNBdXRvKCkpKSB7Ci0gICAgICAgICAgICBjaGls ZC0+c2V0Q2hpbGROZWVkc0xheW91dCh0cnVlLCBNYXJrT25seVRoaXMpOworICAgICAgICAgICAg TGVuZ3RoIGNoaWxkTWFpbkF4aXNMZW5ndGggPSBpc0NvbHVtbkZsb3coKSA/IGNoaWxkLT5zdHls ZSgpLT5sb2dpY2FsV2lkdGgoKSA6IGNoaWxkLT5zdHlsZSgpLT5sb2dpY2FsSGVpZ2h0KCk7Cisg ICAgICAgICAgICBpZiAoY2hpbGRNYWluQXhpc0xlbmd0aC5pc1BlcmNlbnQoKSkKKyAgICAgICAg ICAgICAgICBjaGlsZC0+c2V0Q2hpbGROZWVkc0xheW91dCh0cnVlLCBNYXJrT25seVRoaXMpOwog ICAgICAgICAgICAgY2hpbGQtPmxheW91dElmTmVlZGVkKCk7CiAgICAgICAgIH0KIAo=