[Webkit-unassigned] [Bug 96382] Android's mock scrollbars shows up as a difference in layout test results

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=96382





--- Comment #8 from Peter Beverloo <peter at chromium.org>  2012-09-12 10:40:03 PST ---
(In reply to comment #6)
> How does garden-o-matic distinguish tests to be rebaselined and bugs to be fixed, among the thousands of mismatches?
> 
> 
> The following are a bit off topic:
> 
> Perhaps we can make the pixel results automatically rebaselined, but for crash/timeout/text mismatches, I think we still need to manually check one by one to distinguish bugs from rebaselines. So I think keeping the text expectations as many matching is important to reduce the cost.
> 
> For the scrollbar case, the current code makes the text expectations matching chromium-linux/chromium-win. So does the code for font configurations. For those features, DumpRenderTree doesn't execute the actual code paths of chromium-android. I think we can add several Android-specific test cases for them with the actual paths selected with some Android-specific layout test API. For the other tests the DumpRenderTree code paths are used to reduce the number of mismatches.

We shouldn't be having thousands of mismatches to start with, but rather a good baseline of the expected results. Crashes will show up independently.

With our own larger set of baselines, we can also remove a number of these exceptions and start testing what we use. Fonts would be an exception here due to the shaping and dimensions, unfortunately.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.