Layout Regression on top sites when display: flex is used |
||||||
Issue descriptionChrome Version: Canary 59.0.3035.0 and last few builds Bisect returns https://chromium.googlesource.com/chromium/src/+log/9f04f39d54f599a346c59670f91923f4af05e416..f5a29c6561e3820a5e71e03c85eb7597b797bfd7 OS: Windows 10 What steps will reproduce the problem? (1) Open https://developer.apple.com/reference/uikit/nsunderlinestyle OBSERVE: Bottom of page scrolled off to the far right. (2) Open Twitter DM conversation UI. OBSERVE: Text Edit box scrolled out of view. Of the CLs in range, https://chromium.googlesource.com/chromium/src/+/a90dba1f2dd094e51229dd2e5e059c6b53cee93a seems most suspect?
,
Mar 8 2017
After running "git revert a90dba1f2dd094e51229dd2e5e059c6b53cee93a" on my Canary build, both pages layout correctly.
,
Mar 8 2017
,
Mar 8 2017
Related bug explains more clearly what is going on here ("Elements with display: flex and ::before display: row render content outside content box") and it includes a reduced test case.
Example URL:
http://codepen.io/tcschiller/pen/Qpdbjo
Steps to reproduce the problem:
The above codepen shows the issue. If flex on the element itself or table on the ::before element is removed, content renders correctly.
,
Mar 8 2017
I haven't look at the issue in detail, but it may be related to my change, indeed. In fact, I'm investigating issue #699116 as well, probably caused by the mentioned patch. Perhaps it'd be a good idea to revert it, so I can investigate both issues more carefully.
,
Mar 9 2017
https://crrev.com/2528253003 has been reverted. Could anybody verify the regression is solved now ?
,
Mar 9 2017
After syncing to the latest (f78878b8ba3) and rebuilding Chromium, this is now fixed (verified on Twitter, Apple docs, and the codepen repro)
,
Mar 9 2017
It might be a good idea to land a layout test for this, since it's pretty amazing that we didn't catch this in the CQ for the original change.
,
Mar 10 2017
In any case, I think we should merge https://crrev.com/2740063003 in M59 as soon as possible.
,
Mar 10 2017
,
Mar 10 2017
Sorry about the confusion, no need to merge.
,
Mar 10 2017
I'll mark this issue as FIXED then. Thanks for verifying it.
,
Mar 10 2017
I'll try to define a layout test for this issue as part of the proper fix of issue 667785, which patch has been the responsible of several regressions. I'll try also t define layout test cases based on them. |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by elawrence@chromium.org
, Mar 8 2017Labels: has-Bisect