User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:50.0) Gecko/20100101 Firefox/50.0 Build ID: 20161209095719 Steps to reproduce: 0. Use Firefox on Ubunutu (latest from the Ubuntu repository, Firefox 50.1.0) - did not yet test this on other platforms 1. Create a table width DIVs, based on CSS display: table, table-row, table-cell 2. Gave one of the rows a background color 3. Set text direction: rtl (right-to-left) 4. Resized window very small (width) and then large again (depends on test case) Actual results: The background for the table row is not rendered completely (see screenshot). In a more complex layout (more rows) this also happens without resizing. The problem went worse when opening the developer console. Therefore, I assume it's a performance issue. Please see attached screenshots for example. Expected results: The background should fill the whole table row.

Coukd you just provide a testcase as html file like on jsfiddle or codepen, please.

Ah .. JSFiddle is a good idea for that. Here's the test case: https://jsfiddle.net/xgp50efy/ Please resize the output window small and then large again to see the issue. (Tested with Firefox 50.1.0 under Ubuntu 16.04)

Regression range: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=a38c23ccca0a&tochange=be81b8d6fae9 It could be a regression from bug 1169440. David, what do you think?

Seems much more likely to be a regression from bug 1174700 (also in that range).

Narrowing regression range: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=cd741d3ae78a&tochange=ed293fc9596c

(In reply to David Baron :dbaron: ⌚️UTC-8 from comment #4) > Seems much more likely to be a regression from bug 1174700 (also in that > range). Yes, that'll surely be it. Interesting that the testcase only exhibits the problem when the table cells have position:relative; if I remove that, I can no longer reproduce. Another version of the issue: open the testcase (attachment 8823134 [details]), and use the Dev Tools inspector to look at the style of one of the table cells. Click the checkbox to disable the "position: relative" property on it: no visible change (as expected). Then click it again to re-enable the "position: relative" property: cell background disappears. Slightly resizing the window makes it reappear.

Hi Jonathan, we looked at this in regression triage today. Is this bad enough to get an assignee (or a backout), or leave it for a later fix?

(In reply to David Bolter [:davidb] from comment #7) > Hi Jonathan, we looked at this in regression triage today. Is this bad > enough to get an assignee (or a backout), or leave it for a later fix? It looks like this is limited to a pretty specific (and rare, I guess) combination of circumstances, and it's more of a cosmetic than a crippling failure, such that I don't think a backout is called for (and trying to back out bug 1174700 at this point would carry substantial risk of its own). I wouldn't consider it a top priority, either, though it's certainly a valid bug we should fix.

Too late for firefox 52, mass-wontfix.

Given comment 8 I'm letting this fall off of our weekly platform triage. Would it make a good first bug?

(In reply to Liz Henry (:lizzard) (needinfo? me) from comment #10) > Given comment 8 I'm letting this fall off of our weekly platform triage. > Would it make a good first bug? Probably not; my gut feeling is that it's a layout issue that may be tricky to get right, and will depend on understanding reflow and frame coordinate management in some depth.