Closed Bug 698985 Opened 12 years ago Closed 12 years ago

Near-permaorange Android TEST-UNEXPECTED-FAIL | canvas/image-rendering-test.html | image comparison (==)

Categories

(Core :: Graphics: Canvas2D, defect)

Product:
Core
Component:
Graphics: Canvas2D
Platform:
ARM
Android
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED FIXED
Milestone:
mozilla11

People

(Reporter: philor, Unassigned)

References

Details

(Keywords: intermittent-failure)

Attachments

(2 files)

reftest log
16.30 KB, text/plain
Details
patch
1.85 KB, patch
dholbert
: review+
Details | Diff | Splinter Review
Attached file reftest log — 
https://tbpl.mozilla.org/php/getParsedLog.php?id=7156947&tree=Mozilla-Inbound
Android Tegra 250 mozilla-inbound opt test reftest-1 on 2011-11-01 22:05:54 PDT for push 93625e3af8bc

REFTEST TEST-START | http://10.250.48.212:30102/tests/layout/reftests/canvas/image-rendering-test.html | 2752 / 3583 (76%)
REFTEST TEST-START | http://10.250.48.212:30102/tests/layout/reftests/canvas/image-rendering-ref.html | 2752 / 3583 (76%)
REFTEST TEST-UNEXPECTED-FAIL | http://10.250.48.212:30102/tests/layout/reftests/canvas/image-rendering-test.html | image comparison (==) 
...
REFTEST number of differing pixels: 40000
REFTEST INFO | Saved log: START http://10.250.48.212:30102/tests/layout/reftests/canvas/image-rendering-test.html
REFTEST INFO | Saved log: [CONTENT] OnDocumentLoad triggering AfterOnLoadScripts
REFTEST INFO | Saved log: Initializing canvas snapshot
REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000
REFTEST INFO | Saved log: [CONTENT] RecordResult fired
REFTEST INFO | Saved log: RecordResult fired
REFTEST INFO | Saved log: START http://10.250.48.212:30102/tests/layout/reftests/canvas/image-rendering-ref.html
REFTEST INFO | Saved log: [CONTENT] OnDocumentLoad triggering AfterOnLoadScripts
REFTEST INFO | Saved log: Initializing canvas snapshot
REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000
REFTEST INFO | Saved log: [CONTENT] RecordResult fired
REFTEST INFO | Saved log: RecordResult fired
In the hypothetical case where someone might notice this became permaorange across trees with no code connection this morning: the most likely thing among the releng changes that landed today is splitting reftests into 3 chunks instead of 2, with the probable result that whatever earlier test this was relying on after-effects from to be able to (sometimes) pass is no longer part of the same hunk as this one.
Summary: Intermittent Android TEST-UNEXPECTED-FAIL | canvas/image-rendering-test.html | image comparison (==) → Permaorange Android TEST-UNEXPECTED-FAIL | canvas/image-rendering-test.html | image comparison (==)
Annotated as failing on Android in https://hg.mozilla.org/integration/mozilla-inbound/rev/4bdff61214c0 (not that that will help aurora, or beta, or release).
(In reply to Phil Ringnalda (:philor) from comment #65)
> Annotated as failing on Android in
> https://hg.mozilla.org/integration/mozilla-inbound/rev/4bdff61214c0 (not
> that that will help aurora, or beta, or release).

https://hg.mozilla.org/mozilla-central/rev/4bdff61214c0
Target Milestone: --- → mozilla11
so in the test image, we show a blank white image, and the reference shows a green square surrounded by a red square.

The differences here are:

Test File:
<html reftest-zoom="2">
  <canvas style="image-rendering: -moz-crisp-edges; " id="canvas" width="300" height="300"></canvas>


reference file:
  <canvas id="canvas" width="300" height="300"></canvas>



so why is one of these two causing a failure:
* style="image-rendering: -moz-crisp-edges; "
* reftest-zoom="2"

Unfortunately we have both of these defined in layout/reftests/image/background-image-zoom-1.html and that test seems to be passing just fine in reftest2 chunk.

The big difference is that we are testing a canvas vs html for the style.
Attached patch patch — — Splinter Review 
maybe the zoom happens after the canvas has been drawn on Android. In which case this should fix it.
I can try this patch out a bit later this morning.
odd, I am unable to reproduce this problem on inbound.  I pulled out the skip-if(Android) and couldn't get it to fail.  I have tried running it standalone, in a directory, and as a chunk
also I tested with the patch on this bug and the tests passed just fine.  If we feel this would be a good patch to land and the test would be just as effective, then lets review this and run it through try server.
Sent to try, passed on on Android.
Attachment #575429 - Flags: review?(dholbert)
Attachment #575429 - Flags: review?(dholbert) → review+
Perma, Almostperma, who can tell the difference?
Summary: Permaorange Android TEST-UNEXPECTED-FAIL | canvas/image-rendering-test.html | image comparison (==) → Near-permaorange Android TEST-UNEXPECTED-FAIL | canvas/image-rendering-test.html | image comparison (==)
(In reply to Robert Longson from comment #84)
> pushed https://hg.mozilla.org/integration/mozilla-inbound/rev/d84b129b9936

https://hg.mozilla.org/mozilla-central/rev/d84b129b9936
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
can we land this patch on aurora as well?
Still getting lots of these failures, reopening
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Not on the trunk we aren't, just down in the depths of the release train.
Status: REOPENED → RESOLVED
Closed: 12 years ago → 12 years ago
Resolution: --- → FIXED
Since this patch is a test-only change, can we land this on mozilla-beta to get rid of the permaorange there?
Comment on attachment 575429 [details] [diff] [review]
patch

Does this patch apply to the ESR as well? If so, it would be good to take the fix there as well to prevent the perma-orange.
(In reply to Alex Keybl [:akeybl] from comment #291)
> Does this patch apply to the ESR as well? If so, it would be good to take
> the fix there as well to prevent the perma-orange.

Thanks for the heads-up on that.  Indeed, the patch does apply to ESR.  (minus the removal of the "random" annotation in reftest.list, but that's fine -- ESR just never got that annotation in the first place (which is why it's still turning the tree orange).)

I pushed longsonr's patch to ESR, with a=tests-only:
  https://hg.mozilla.org/releases/mozilla-esr10/rev/7a5868ec89c9
Whiteboard: [orange]
You need to log in before you can comment on or make changes to this bug.