Bug 85772

Summary: [chromium] REGRESSION(r107389) Visible line artifacts on some JPEG images
Product: WebKit Reporter: noel gordon <noel.gordon>
Component: New BugsAssignee: noel gordon <noel.gordon>
Status: RESOLVED FIXED    
Severity: Normal CC: abarth, dpranke, eric, pkasting, tkent, webkit.review.bot
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: Unspecified   
OS: Unspecified   
URL: http://www.polepositiongroup.com/img/banners/bg_006.jpg
Bug Depends on:    
Bug Blocks: 78323    
Attachments:
Description Flags
chrome-20.0.1123.4-jpeg-artifacts.png
none
firefox-12.0-jpeg-artifacts.png
none
firefox-9.0-jpeg.png
none
Patch
none
Patch
none
Patch for landing none

Description noel gordon 2012-05-06 22:14:50 PDT
Created attachment 140473 [details]
chrome-20.0.1123.4-jpeg-artifacts.png

http://www.polepositiongroup.com/img/banners/bg_006.jpg image shows visible horizontal and vertical line artifacts.

Chrome 20.0.1124.4 zoomed in image capture attached, shows clear vertical lines artifacts.
Comment 1 noel gordon 2012-05-06 22:17:31 PDT
Created attachment 140474 [details]
firefox-12.0-jpeg-artifacts.png
Comment 2 noel gordon 2012-05-06 22:22:23 PDT
Compared Firefox 12.0 Mac rendering, I see the same vertical & horizontal artifacts (picture above).

I _don't_ see the artifacts using Firefox 9.0 Mac, so the issue seems to be a regression in libjpeg-turbo that manifests in both Firefox 12, and Chrome 20 http://crbug.com/122951
Comment 3 noel gordon 2012-05-06 22:35:39 PDT
Created attachment 140477 [details]
firefox-9.0-jpeg.png

No artifacts show up on Firefox 9.0 Mac.
Comment 4 noel gordon 2012-05-08 22:16:55 PDT
Filed a mozilla bug: https://bugzilla.mozilla.org/show_bug.cgi?id=753235
Comment 5 noel gordon 2012-05-09 23:26:07 PDT
Looked at Firefox 12.0 again today and I don't see artifacts.  Firefox 9.0, 12.0 render fine.
Comment 6 noel gordon 2012-05-10 00:30:12 PDT
Next punt IFAST vs ISLOW DCT decodes.  No artifacts with ISLOW, so backing that out, patch coming ...
Comment 7 noel gordon 2012-05-10 00:33:34 PDT
Created attachment 141100 [details]
Patch
Comment 8 Kent Tamura 2012-05-10 00:42:58 PDT
Comment on attachment 141100 [details]
Patch

rubber-stamped.
Comment 9 noel gordon 2012-05-10 00:55:35 PDT
Thank you, and a question: do rebaselines seem slow from your location?
Comment 10 Kent Tamura 2012-05-10 00:57:43 PDT
(In reply to comment #9)
> Thank you, and a question: do rebaselines seem slow from your location?

Both of "webkit-patch rebaseline-expectatinos" and garden-o-matic are very slow in Tokyo.
Comment 11 noel gordon 2012-05-10 01:03:37 PDT
Same from Sydney.  Over the last three weeks (it seems to me) these tools became noticeably slower.
Comment 12 Eric Seidel (no email) 2012-05-10 01:08:59 PDT
You might try running with -v, to see what those tools are doing.  I'm not pariticularly famiiliar with either.
Comment 13 noel gordon 2012-05-10 01:18:36 PDT
Good idea.
Comment 14 noel gordon 2012-05-10 01:20:03 PDT
(In reply to comment #8)
> (From update of attachment 141100 [details])
> rubber-stamped.

Need to add = IMAGE expectations for some additional tests

  fast/repaint/block-layout-inline-children-replaced.html
  tables/mozilla_expected_failures/bugs/bug85016.html
Comment 15 noel gordon 2012-05-10 01:29:20 PDT
Created attachment 141107 [details]
Patch
Comment 16 noel gordon 2012-05-10 01:38:51 PDT
(In reply to comment #4)
> Filed a mozilla bug: https://bugzilla.mozilla.org/show_bug.cgi?id=753235

Closed.
Comment 17 noel gordon 2012-05-10 02:44:19 PDT
Created attachment 141120 [details]
Patch for landing
Comment 18 WebKit Review Bot 2012-05-10 06:15:06 PDT
Comment on attachment 141120 [details]
Patch for landing

Clearing flags on attachment: 141120

Committed r116636: <http://trac.webkit.org/changeset/116636>
Comment 19 WebKit Review Bot 2012-05-10 06:15:15 PDT
All reviewed patches have been landed.  Closing bug.
Comment 20 noel gordon 2012-05-10 20:55:27 PDT
(In reply to comment #12)
> You might try running with -v, to see what those tools are doing.  I'm not pariticularly famiiliar with either.

Filed bug 86171 with a -v trace of a rebaseline of one test: total time was 21m 14sec :/