389496 - Bad placement of absolute span inside of fixed span with display:inline-table

Status: NEW

(Core :: Layout, defect)

Description

[Daniel Holbert [:dholbert]]

Found this while editing testcases for bug 363729.

Steps to reproduce:
-------------------
 - Load testcase.
 - Resize window.

Observed behavior:
------------------
 - Initial pageload shows "Inner" at coordinate (0,0), to upper-left of and on top of "Outer".
 - The window resize triggers a reflow which moves "Inner" to right of "Outer"

Expected behavior:
------------------
 - Initial pageload should place "Inner" to the right of "Outer".  
  (This would match FF 2.0 behavior, and it would be consistent with Trunk behavior after the window-resize)

Comment 1

[Daniel Holbert [:dholbert]]

Attachment: testcase

Comment 2

[dolphinling]

regression based on comment 0, qawanted for regression range

Comment 3

[Daniel Holbert [:dholbert]]

Attachment: testcase using explicit table (instead of "display: table-inline")

Comment 4

[Daniel Holbert [:dholbert]]

Attachment: testcase using explicit table #2 (instead of "display: table-inline")

(set cellspacing / cellpadding to 0 in this version)

Comment 5

[Daniel Holbert [:dholbert]]

(In reply to comment #0)
>   (This would match FF 2.0 behavior, and it would be consistent with Trunk
> behavior after the window-resize)
> 

Oops as it turns out, Firefox 2.0 doesn't support "display: table-inline", and trunk does, so they handled the first testcase differently for that reason.

The new testcase (attachment 274223 [details]), using an explicit table, exhibits the original "observed behavior" from Comment 0 in both trunk AND branch.  (inner span jumps during the first window resize) 

I'm assuming the original "Expected Behavior" is still correct, because that that's how both IE and Opera handle this testcase. (And even if they're wrong, we're definitely not right with our jumping span.)

Comment 6

[Daniel Holbert [:dholbert]]

Removing regression and qawanted keywords, because it's not actually a regression and so we don't need to get a regression range. (as described in comment #5.)