Bug 86712 – Layout Test tables/mozilla/marvin/backgr_position-table.html is failing

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Hajime Morrita]

Following tests are failing. The border painting was changed.

tables/mozilla/marvin/backgr_position-table.html
tables/mozilla_expected_failures/marvin/backgr_border-table-row.html

See:
http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=tables%2Fmozilla%2Fmarvin%2Fbackgr_position-table.html
http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=tables%2Fmozilla_expected_failures%2Fmarvin%2Fbackgr_border-table-row.html

Possible affecting change is http://trac.webkit.org/changeset/117271

I'm not sure if these just needs rebaseline or are broken.
Marking as fail for now.

[Arpita Bahuguna]

I will check on this and update shortly.

[Arpita Bahuguna]

(In reply to comment #0)
> Following tests are failing. The border painting was changed.
> 
> tables/mozilla/marvin/backgr_position-table.html
> tables/mozilla_expected_failures/marvin/backgr_border-table-row.html

Hi Hajime,
I verified the testcases with and without my changes on mac (v 10.7.3) webkit - r117432.
Have uploaded the results for the same.

> 
> See:
> http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=tables%2Fmozilla%2Fmarvin%2Fbackgr_position-table.html
> http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=tables%2Fmozilla_expected_failures%2
Fmarvin%2Fbackgr_border-table-row.html

The results that I get on my machine seem to be different from the ones above. Probably am missing out on some fonts.


> 
> Possible affecting change is http://trac.webkit.org/changeset/117271
> 

My fix is specifically pertaining to collapsed-borders set as "collapsed". The first table for which the image shows the difference has collapsed-borders set as "separate" and the change itself is not executed while running the testcases.
Probably the issue could be due to fonts (since the widths of blocks containing texts are shown to be different).
I also verified the renderTree dump for the block containing the diff border and found the values to be the same on both actual and expected results.
 
> I'm not sure if these just needs rebaseline or are broken.
> Marking as fail for now.

Rebaselining could perhaps fix the issue? 
Please let me know how to further proceed with this.

[Arpita Bahuguna]

Created attachment 142469 [details]
Layout test results with the fix - 117271

[Arpita Bahuguna]

Created attachment 142470 [details]
Layout test results without the fix

[Julien Chaffraix]

> Rebaselining could perhaps fix the issue?

Please don't rebaseline. One of the new baseline is an obvious regression (tables/mozilla_expected_failures/marvin/backgr_border-table-row.html). The other one would need to be investigated as we don't see if the bottom border is painted.

I agree with Arpita, I don't think r117271 is the culprit as the pixel diffs are on border: separate tables and the change is specific to border: collapse.

Furthermore, the test only fails on Mac and is starting to be passing again on some bots. Looking at the blamelist, there is a Skia roll (http://src.chromium.org/viewvc/chrome?view=rev&revision=137601) which could be our culprit.

[Peter Kasting]

The two tests named in comment 0 are now passing, looks to have been a Skia issue fixed in http://src.chromium.org/viewvc/chrome?view=rev&revision=137651 .

A number of other tests were added to this bug in the expectations file but seem unrelated; I'm going to split those off and close this one.

Expectations removed in r117492.

[Hajime Morrita]

(In reply to comment #6)
> The two tests named in comment 0 are now passing, looks to have been a Skia issue fixed in http://src.chromium.org/viewvc/chrome?view=rev&revision=137651 .
> 
> A number of other tests were added to this bug in the expectations file but seem unrelated; I'm going to split those off and close this one.
> 
> Expectations removed in r117492.

Thanks for the investigation folks, 
I'm sorry for the false alarm, Arpita. I overlooked the Skia roll...