Bug 69956 – baseline images for fast/repaint/japanese-rl-selection-repaint-in-regions seem wrong

• Component: Tools / Tests
• Status: RESOLVED FIXED

I notice that this one layout test has very different expectations than I see in shipping Chrome and Safari.

I see that a few different people (CCed here) have checked in baselines for this test in recent weeks.  Any ideas?

These three versions look about the same as each other, but different from what I see in shipping Chrome/Safari:
http://trac.webkit.org/browser/trunk/LayoutTests/platform/mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png
http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-cg-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png
http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-cg-mac-leopard/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png

This one looks a little different from the above, but also different from what I see in shipping Chrome/Safari:
http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png

I have also attached screen shots of what I see when viewing http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-repaint-in-regions.html in these browsers...

These two look similar to each other but quite different from all the expectations above:
Chrome 14.0.835.202
Safari 5.1 (6534.50)

This one looks like the expectations above (not surprising) but very different from shipping Chrome.
Chromium Canary 16.0.905.0

Created attachment 110729 [details]
chrome 14.0.835.202

Created attachment 110730 [details]
chromium canary 16.0.905.0

Created attachment 110731 [details]
safari 5.1

[Ryosuke Niwa]

Clearly ToT result is correct.

[Ryosuke Niwa]

(In reply to comment #1)
> Created an attachment (id=110729) [details]
> chrome 14.0.835.202

(In reply to comment #3)
> Created an attachment (id=110731) [details]
> safari 5.1

They both look wrong.

[Ryosuke Niwa]

(In reply to comment #0)
> These three versions look about the same as each other, but different from what I see in shipping Chrome/Safari:
> http://trac.webkit.org/browser/trunk/LayoutTests/platform/mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png
> http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-cg-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png
> http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-cg-mac-leopard/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png
> 
> This one looks a little different from the above, but also different from what I see in shipping Chrome/Safari:
> http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png

As far as I can tell, they all look the same although they repaint slightly different regions.

Created attachment 110881 [details]
chrome canary vs shipping chrome

The attached PDF shows Chrome Canary on the top and Shipping Chrome at the bottom.

The three columns are these 3 pages, as rendered in each browser:
http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-clear.html
http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-repaint-in-regions.html
http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-repaint.html

Pages #1 and #3 render the same in both browsers.  Why does page #2 render so differently?

[Ryosuke Niwa]

Why isn't hyatt cc-ed on this bug?

[Ryosuke Niwa]

(In reply to comment #7)
> The three columns are these 3 pages, as rendered in each browser:
> http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-clear.html
> http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-repaint-in-regions.html
> http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-repaint.html
> 
> Pages #1 and #3 render the same in both browsers.  Why does page #2 render so differently?

Because it pulls the content out of the original div and put in an absolutely positioned box:

<div style="-webkit-flow:'thread'">
せっかく見つけたすばらしい記事がどこにあったか忘れてしまった経験はありますかならタイトルとアドレスだけでなく、訪問したウェブページのコンテンツからも検索することができます。せっかく見つけたすばらしい記事がどこにあったか忘れてしまった経験はありますか ならタイトルとアドレスだけでなく、訪問したウェブページのコンテンツからも検索することができます。訪問したウェブページのコンテンツからも検索することができます。せっかく見つけたすばらしい記事がどこにあったか忘れてしまった経験はありますか ならタイトルとアドレスだけでなく、訪問
</div>
<div style="content:-webkit-from-flow('thread'); position:absolute;right:100px; top:100px; border:2px solid black; width:400px; height:400px"></div>

(In reply to comment #9)
> (In reply to comment #7)
> > The three columns are these 3 pages, as rendered in each browser:
> > http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-clear.html
> > http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-repaint-in-regions.html
> > http://trac.webkit.org/export/93144/trunk/LayoutTests/fast/repaint/japanese-rl-selection-repaint.html
> > 
> > Pages #1 and #3 render the same in both browsers.  Why does page #2 render so differently?
> 
> Because it pulls the content out of the original div and put in an absolutely positioned box:

OK, Ryosuke, thanks.

Am I correct in assuming that this expected image is correct:
http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-cg-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png

As opposed to this one, which is almost the same but the selection of the final line is clipped differently:
http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png

[Ryosuke Niwa]

(In reply to comment #10)
> OK, Ryosuke, thanks.
> 
> Am I correct in assuming that this expected image is correct:
> http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-cg-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png
> 
> As opposed to this one, which is almost the same but the selection of the final line is clipped differently:
> http://trac.webkit.org/browser/trunk/LayoutTests/platform/chromium-mac/fast/repaint/japanese-rl-selection-repaint-in-regions-expected.png

Skia one looks wrong. It shouldn't be truncating lines of text like that as far as I know.

New chromium-mac baseline image committed within http://trac.webkit.org/changeset/97775 .  I think we can close this out now.