Bug 225952 – [GTK][WebInspector] Value of 'imageSmoothingQuality' is 'medium' for platforms not using CG

• Component: New Bugs
• Status: RESOLVED FIXED

[Diego Pino]

[GTK][WebInspector] Value of 'imageSmoothingQuality' is 'medium' for platforms not using CG

[Diego Pino]

Created attachment 429015 [details]
Patch

[Diego Pino]

Right now several tests in 'inspector/canvas' are failing because of a wrong check of the 'imageSmoothingQuality' value. This value is 'low' for CG platforms and 'medium' otherwise [1]. The tests that are failing are the following:

  inspector/canvas/recording-2d-frameCount.html
  inspector/canvas/recording-2d-full.html
  inspector/canvas/recording-2d-memoryLimit.html
  inspector/canvas/recording-2d-saves.html
  inspector/canvas/recording-html-2d.html

The patch sets the default value accordingly and emits new specific baselines for GTK platform. Maybe emitting new baselines can be saved if the check of 'imageSmoothingQuality' value is done in a different way.

[1] https://webkit-search.igalia.com/webkit/source/Source/WebCore/html/canvas/CanvasRenderingContext2DBase.cpp#86

[Sam Weinig]

Perhaps we should instead update non-CG platforms to use low as well, as that is what the spec says it should be these days: https://html.spec.whatwg.org/multipage/canvas.html#canvasimagesmoothing

[Diego Pino]

Created attachment 429142 [details]
Patch

[Diego Pino]

Created attachment 429148 [details]
Patch

[Sam Weinig]

Comment on attachment 429148 [details]
Patch

Thanks!

[EWS]

Committed r277896 (238031@main): <https://commits.webkit.org/238031@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 429148 [details].

[Radar WebKit Bug Importer]

<rdar://problem/78333302>

[Tim Horton]

FWIW this broke 

fast/canvas/canvas-imageSmoothingEnabled.html
fast/canvas/canvas-imageSmoothingQuality.html

on gtk

[Tim Horton]

Also why the [WebInspector] prefix? This is a wide-ranging change :)

[Tim Horton]

From my debugging before I found this bug yesterday, it kind of seemed like Cairo's "low quality" really means "not at all". Maybe that's why it was medium before?

[Diego Pino]

(In reply to Tim Horton from comment #9)
> FWIW this broke 
> 
> fast/canvas/canvas-imageSmoothingEnabled.html
> fast/canvas/canvas-imageSmoothingQuality.html
> 
> on gtk

Right, fast/canvas/canvas-imageSmoothingEnabled.html has regressed:

-PASS left_of_center_pixel.data[0] !== 0 is true
-PASS left_of_center_pixel.data[1] !== 0 is true
-PASS left_of_center_pixel.data[2] !== 0 is true
+FAIL left_of_center_pixel.data[0] !== 0 should be true. Was false.
+FAIL left_of_center_pixel.data[1] !== 0 should be true. Was false.
+FAIL left_of_center_pixel.data[2] !== 0 should be true. Was false.

fast/canvas/canvas-imageSmoothingQuality.html has its own baseline for GTK. The test has now one test that went from FAIL to PASS and viceversa. I think updating the baseline is the right thing to do.

--- /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/fast/canvas/canvas-imageSmoothingQuality-expected.txt
+++ /home/buildbot/worker/gtk-linux-64-release-tests/build/layout-test-results/fast/canvas/canvas-imageSmoothingQuality-actual.txt
@@ -8,14 +8,14 @@
 PASS mediumContext.imageSmoothingQuality is 'medium'
 PASS highContext.imageSmoothingQuality is 'high'
 
-PASS lowData is not mediumData
+FAIL lowData should not be {"0":0,"1":0,"2":0,"3":128}.
 PASS mediumData is not highData
 PASS lowData is not highData
 
 PASS sampleAlpha(lowData) is >= sampleAlpha(mediumData)
 PASS sampleAlpha(mediumData) is >= sampleAlpha(highData)
 
-FAIL defaultContext.imageSmoothingQuality should be low. Was medium.
+PASS defaultContext.imageSmoothingQuality is 'low'
 
 PASS highContext.imageSmoothingEnabled = false; highContext.imageSmoothingQuality is 'high'
 PASS highContext.imageSmoothingQuality = 'medium'; highContext.imageSmoothingQuality is 'medium'

[Diego Pino]

(In reply to Tim Horton from comment #10)
> Also why the [WebInspector] prefix? This is a wide-ranging change :)

The first patch changed LayoutTests/inspector/canvas/recording-2d-saves.html to produce different imageSmoothingQuality values depending on platform. Since it only targeted a WebInspector test adding the WebInspector label made sense. Then, in the review process it was pointed out imageSmoothingQuality default value should be Low as defined by specification, but the original WebInspector label remained.

[Diego Pino]

(In reply to Tim Horton from comment #11)
> From my debugging before I found this bug yesterday, it kind of seemed like
> Cairo's "low quality" really means "not at all". Maybe that's why it was
> medium before?

I will file a new bug to keep track of the tests that have regressed after this change:

fast/canvas/canvas-imageSmoothingEnabled.html
fast/canvas/canvas-imageSmoothingQuality.html

From my POV, and without looking into this issue yet, it makes sense to me to keep imageSmoothingQuality default value as 'Low' and try to solve this issue at Cairo level.

[Diego Pino]

*** Bug 223301 has been marked as a duplicate of this bug. ***