Bug 22824 – REGRESSION (r39206): 3 SVG pixel tests are failing on the Leopard pixel bot since r39206

• Component: New Bugs
• Status: RESOLVED FIXED

[Pierre-Olivier Latour]

3 SVG pixel tests are failing on the Leopard pixel bot since build 884:

http://build.webkit.org/builders/trunk-mac-intel-pixel/builds/883 (r39205)
http://build.webkit.org/builders/trunk-mac-intel-pixel/builds/884 (r39206)

svg/W3C-SVG-1.1/text-tselect-02-f.svg	expected	actual	diffs	expected image	image diffs	1.21%
svg/custom/image-small-width-height.svg	expected	actual	diffs	expected image	image diffs	0.28%
svg/text/text-tselect-02-f.svg	expected	actual	diffs	expected image	image diffs	1.23%

Here's the change: http://trac.webkit.org/changeset/39206

[Pierre-Olivier Latour]

I checked in new results so that the bots are green again:

Sending        LayoutTests/ChangeLog
Sending        LayoutTests/platform/mac/svg/W3C-SVG-1.1/text-tselect-02-f-expected.checksum
Sending        LayoutTests/platform/mac/svg/W3C-SVG-1.1/text-tselect-02-f-expected.png
Sending        LayoutTests/platform/mac/svg/W3C-SVG-1.1/text-tselect-02-f-expected.txt
Sending        LayoutTests/platform/mac/svg/custom/image-small-width-height-expected.checksum
Sending        LayoutTests/platform/mac/svg/custom/image-small-width-height-expected.png
Sending        LayoutTests/platform/mac/svg/custom/image-small-width-height-expected.txt
Sending        LayoutTests/platform/mac/svg/text/text-tselect-02-f-expected.checksum
Sending        LayoutTests/platform/mac/svg/text/text-tselect-02-f-expected.png
Sending        LayoutTests/platform/mac/svg/text/text-tselect-02-f-expected.txt
Transmitting file data ..........
Committed revision 39254.

[Holger Freyther]

What am I supposed to do?

(In reply to comment #2)
> What am I supposed to do?
> 

Maybe change the single instance of ENABLE(FONT_FAST_PATH) in your patch to USE(FONT_FAST_PATH) as a start.

[Holger Freyther]

Created attachment 26002 [details]
Attempt to fix the regression

Ouch. The patch was changed to use USE(...) instead of ENABLE(...) and this one item was not changed. When looking at the patch I didn't see the wrong #if. I think the other part of the patch is right. I'm currently not on OSX and I'm not able to run the pixel tests.

The change from:
  if(canUse)
    simple
  else
    complex
}

to
  if (canUse)
     return simple;
   return complex
}

 
looks okay and is matching other parts of the code...

[Holger Freyther]

Landed in r39277. The mentioned tests work on trunk-mac-intel-pixel. And on trunk-mac-ppc-pixel the number of pixel failures is down from 4 to 2. Ironically when searching for "-> failed" in the log I only see one unrelated failure.

I assume this issue can be resolved as fixed?

Fixed. Thanks!