Add some canvas tests to test_expectations to prepare for changes to use sw-rasterized paths in skia-gpu
Created attachment 122793 [details] Patch
These tests break in DRT when using the sw rasterizer for paths. I noticed that the downstream test_expectations is not respected for --platform chromium-gpu so instead I'm modifying upstream in advance. Is that a known issue or should I file a chromium bug?
(In reply to comment #2) > These tests break in DRT when using the sw rasterizer for paths. I noticed that the downstream test_expectations is not respected for --platform chromium-gpu so instead I'm modifying upstream in advance. Is that a known issue or should I file a chromium bug?
(In reply to comment #3) > (In reply to comment #2) > > These tests break in DRT when using the sw rasterizer for paths. I noticed that the downstream test_expectations is not respected for --platform chromium-gpu so instead I'm modifying upstream in advance. Is that a known issue or should I file a chromium bug? It's working as intended. It is designed to make the WebKit gardener's job easier, since they can tell more easily just by looking at the canaries/trybots if the upstream test_expectations are correct (without being masked by downstream t_e changes). It makes the life of the Skia gardener more difficult, though. Our usual policy has been to land upstream suppressions preemptively, as you're doing here.
Comment on attachment 122793 [details] Patch OK. r=me
(In reply to comment #4) > (In reply to comment #3) > > (In reply to comment #2) > > > These tests break in DRT when using the sw rasterizer for paths. I noticed that the downstream test_expectations is not respected for --platform chromium-gpu so instead I'm modifying upstream in advance. Is that a known issue or should I file a chromium bug? > > It's working as intended. It is designed to make the WebKit gardener's job easier, since they can tell more easily just by looking at the canaries/trybots if the upstream test_expectations are correct (without being masked by downstream t_e changes). It makes the life of the Skia gardener more difficult, though. > > Our usual policy has been to land upstream suppressions preemptively, as you're doing here. Got it. Thanks for the clarification!
Comment on attachment 122793 [details] Patch Clearing flags on attachment: 122793 Committed r105191: <http://trac.webkit.org/changeset/105191>
All reviewed patches have been landed. Closing bug.
This patch had bad (duplicate) expectations for fast/canvas/image-object-in-canvas.html and fast/canvas/image-pattern-rotate.html. Please lint test_expectation.txt changes by running: new-run-webkit-tests --lint-test-files and fix any errors it reports before landing
Reopening to attach new patch.
Created attachment 122923 [details] Patch
So sorry about the conflicts. The new patch adds the conflicting fast/canvas tests back to all GPU configs and removes conflicting LEOPARD GPU entries. Also, the four philip tests fail on mac so I've removed WIN LINUX tags. It passes the lint checker.
Comment on attachment 122923 [details] Patch Looks good. r=me
Comment on attachment 122923 [details] Patch Clearing flags on attachment: 122923 Committed r105320: <http://trac.webkit.org/changeset/105320>
Created attachment 123132 [details] Patch
There is one more test to add to this list: arc360.html. The latest patch adds it to test_expectations.
Comment on attachment 123132 [details] Patch OK. r=me
Comment on attachment 123132 [details] Patch Clearing flags on attachment: 123132 Committed r105455: <http://trac.webkit.org/changeset/105455>