Bug 78021 – layoutTestController.display() is flaky for SVG tests

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Nikolas Zimmermann]

layoutTestController.display() is flaky for SVG tests. Reduction:
new-run-webkit-tests --tolerance 0 -p --child-processes=1 svg/W3C-SVG-1.1(-SE)/any-test.svg svg/filters/animate-fill.svg
There's a hack in DRT to produce 480x360 expected.pngs for any test starting with "svg/W3C-SVG-1.1" - the resizing of the web view is done when dump()ing the render tree - _after_ layoutTestController.display() was called and the repaint test executed. This makes the grey overlay vanish again.
The first test that tests repainting, running right after a svg/W3C-SVG-1.1/xxxx test case will fill. Uploading a simple fix for Mac DRT soon.

[Nikolas Zimmermann]

Created attachment 125906 [details]
Patch for mac

Comment on attachment 125906 [details]
Patch for mac

View in context: https://bugs.webkit.org/attachment.cgi?id=125906&action=review

> Tools/ChangeLog:11
> +        hack that forces 480x3600 instead of 800x600 pixel test dumps for any test starting with

Typo: 3600

[Nikolas Zimmermann]

Committed r106983: <http://trac.webkit.org/changeset/106983>

[Adam Roben (:aroben)]

I believe DRT/win has extremely similar code. Any chance you'll fix that implementation too?

[Nikolas Zimmermann]

(In reply to comment #4)
> I believe DRT/win has extremely similar code. Any chance you'll fix that implementation too?
Reopning this bug as reminder. I've checked DRT code for all platforms (efl/gtk/cr/..) - those ports all resize the view before running the test, so they don't suffer from the problem, so really only win/ is left to fix. Will upload a patch soon.

[Nikolas Zimmermann]

Created attachment 126275 [details]
Patch for win

[Early Warning System Bot]

Comment on attachment 126275 [details]
Patch for win

Attachment 126275 [details] did not pass qt-ews (qt):
Output: http://queues.webkit.org/results/11473188

[Nikolas Zimmermann]

The Qt build failure is unrelated.

[Adam Roben (:aroben)]

Comment on attachment 126275 [details]
Patch for win

View in context: https://bugs.webkit.org/attachment.cgi?id=126275&action=review

> Tools/DumpRenderTree/win/DumpRenderTree.cpp:726
> +        ::InvalidateRect(webViewWindow, 0, TRUE);
> +        ::SendMessage(webViewWindow, WM_PAINT, 0, 0);

Why do we need to do this even for dumpAsText tests?

[Eric Seidel (no email)]

Comment on attachment 125906 [details]
Patch for mac

Cleared Dan Bernstein's review+ from obsolete attachment 125906 [details] so that this bug does not appear in http://webkit.org/pending-commit.

[Nikolas Zimmermann]

(In reply to comment #9)
> Why do we need to do this even for dumpAsText tests?
Not sure, I just refactored it - it's as it used to be. It looks suspicious though - I don't have a win machine to test though. To stay on the safe-side we could leave it as-is, but I don't know if it works, I couldn't try it.

[Nikolas Zimmermann]

Anyone?

[Nikolas Zimmermann]

Comment on attachment 126275 [details]
Patch for win

Clearing flags on attachment: 126275

Committed r108054: <http://trac.webkit.org/changeset/108054>

[Nikolas Zimmermann]

All reviewed patches have been landed.  Closing bug.