Bug 64675 – [SVG] [Chromium] zero-path-square-cap-rendering.svg is failing

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

[Ryosuke Niwa]

Two tests added by http://trac.webkit.org/changeset/91125/ are failing on Chromium Linux:

Results:
http://build.webkit.org/results/Chromium%20Linux%20Release%20(Tests)/r91171%20(21141)/svg/W3C-SVG-1.1-SE/painting-control-04-f-diffs.html
http://build.webkit.org/results/Chromium%20Linux%20Release%20(Tests)/r91171%20(21141)/svg/custom/zero-path-square-cap-rendering-diffs.html

Flakiness dashboard:
http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=svg%2FW3C-SVG-1.1-SE%2Fpainting-control-04-f.svg%2Csvg%2Fcustom%2Fzero-path-square-cap-rendering.svg&revision=91126&group=%40ToT%20-%20webkit.org

[Ryosuke Niwa]

They are also failing on Chromium Windows:

http://build.webkit.org/results/Chromium%20Linux%20Release%20(Tests)/r91171%20(21141)/svg/W3C-SVG-1.1-SE/painting-control-04-f-diffs.html

http://build.webkit.org/results/Chromium%20Linux%20Release%20(Tests)/r91171%20(21141)/svg/custom/zero-path-square-cap-rendering-diffs.html

[Rob Buis]

Hi,

(In reply to comment #1)
> They are also failing on Chromium Windows:
> 
> http://build.webkit.org/results/Chromium%20Linux%20Release%20(Tests)/r91171%20(21141)/svg/W3C-SVG-1.1-SE/painting-control-04-f-diffs.html
> 
> http://build.webkit.org/results/Chromium%20Linux%20Release%20(Tests)/r91171%20(21141)/svg/custom/zero-path-square-cap-rendering-diffs.html

This should have improved after r91191.
Cheers,

Rob.

[Ryosuke Niwa]

Nope.  They're still failing :(  This might be Skia related.

[Ryosuke Niwa]

Committed r91211: <http://trac.webkit.org/changeset/91211>

[Ryosuke Niwa]

svg/W3C-SVG-1.1-SE/painting-control-04-f.svg is now passing but zero-path-square-cap-rendering.svg is not.

[Rob Buis]

Hi,

(In reply to comment #5)
> svg/W3C-SVG-1.1-SE/painting-control-04-f.svg is now passing but zero-path-square-cap-rendering.svg is not.

For anyone looking into the problem on chromium, one thing to check first is check that the m_fillBoundingBox width and height is really zero. This comes from Path which is platform specific. So, the question is what Skia returns for bbox of a path "M 100 100 Z", if width/height is not exactly zero, the new special handling of zero-length paths will not kick in.
Cheers,

Rob.

[Stephen Chenney]

This is Skia 398: http://code.google.com/p/skia/issues/detail?id=398

Skia throws away the closing "z" in "m 50 50 z" because it considers it redundant. This then causes the Skia SkPath::isEmpty call to fail because Skia considers a path "m 50 50" to be empty.

This makes it impossible to correctly implement SVG zero-length linecaps for such paths.

[Stephen Chenney]

This will be fixed with 71820, I hope.

[Stephen Chenney]

Changes r105878 and r105921 fix the issue.