1343370 - [css-flex] {inc} Abspos flex child (in relpos flex container) gets wrong static position, after dynamic modification to container

Status: RESOLVED FIXED

(Core :: Layout: Positioned, defect, P3)

Description

[Daniel Holbert [:dholbert]]

STR:
 1. Load attached testcase.
 2. (Optional) Click the button a few times.

EXPECTED RESULTS:
The word "End" should always be snapped to the bottom right of the black-bordered container.

ACTUAL RESULTS:
The word "End" is initially sticking off the top left edge of the container -- and if I click the button to make further tweaks, it never gets to the bottom right edge.  It always seems to be positioned using the container's *previous* size. (IIRC we use the container's GetRect() for abspos positioning, and in this case, the container must not have updated its rect with its final size when we position the child.)

I believe this is a regression caused by bug 1269046.

(Though the EXPECTED RESULTS before that bug landed are a bit different, due to targeting different spec language. Back then, the expected results are that the child should be just outside of the black container at the bottom right corner -- i.e. the 0-sized placeholder was snapped to the bottom right corner.  Now, we snap the child itself to the bottom right corner.)

Comment 1

[Daniel Holbert [:dholbert]]

Attachment: testcase 1

Comment 2

[Daniel Holbert [:dholbert]]

(I'm a bit surprised we haven't noticed any bug reports about this. Hopefully that means not many sites are affected by this issue.)

I *think* this only causes problems when the flex container is the abspos-containing-block, *as well as* the parent, for the abspos thing.  And even then, it'll be fixed if we happen to trigger an extra reflow.

Comment 3

[Daniel Holbert [:dholbert]]

(BTW, I discovered this when testing the sample code on bug 1341560, to be sure that bug was actually fixed by its dupe-target. Under some circumstances, the test code there triggers the issue I'm describing here.)

Comment 4

[yisibl]

Attachment: testcase2.html

The parent element hover will also trigger this bug.

Comment 5

[Astley Chen (inactive)]

It seems too late to fix on 52 as it's spinning RC build now. Daniel, if it's not the case, please correct me. Thanks.

Comment 6

[yisibl]

Attachment: flex-bug-firefox.html

Changing the position value of the flex item can also have a similar problem, and I find that inline-flex is not aligned on the vertical baseline, this is also the bug?

Comment 7

[Liz Henry (:lizzard) (relman/hg->git project)]

Too late for 53, we could still potentially take a patch for 54 though.

Comment 8

[Gerry Chang [:gchang]]

Too late for 54, let's see if we can fix it in 55.

Comment 9

[Ryan VanderMeulen [:RyanVM]]

Fixed by bug 1386654. Is there a need to land a test from this bug still or are we good to go with the ones landed there?