981479 - Numerous layout reftest failures on Windows Server 2012 EC2 instance

Status: NEW

(Core :: Layout, defect, P5)

Description

[(No longer employed by Mozilla) Aaron Klotz]

These tests are run over RDP using the RemoteFX software GPU.

Comment 1

[(No longer employed by Mozilla) Aaron Klotz]

Attachment: Reftest log file

Comment 2

[Mats Palmgren (inactive)]

Most of the failures have "max difference: 1", some have 2, one has 3.  Those can be "fixed"
by specifying some fuzzyness in the manifest file.

One test, layout/reftests/bugs/759036-1.html, is an unexpected-pass.  It's manifest line reads:
fuzzy-if(true,1,19) fails-if(d2d) random-if(Android&&AndroidVersion<15) == 759036-1.html 759036-1-ref.html
I guess this platform has "d2d" being true, should it?

The remaining (4) failures seems like real failures:
content/canvas/test/reftest/webgl-clear-test.html?native-gl
content/canvas/test/reftest/webgl-orientation-test.html?native-gl
content/canvas/test/reftest/webgl-color-test.html?native-gl
content/canvas/test/reftest/webgl-color-test.html?native-gl&16bpp
where the test screenshot simply displays the text "WebGL failed".

It might be worth filing a separate (Graphics) bug to investigate the WebGL part;
the other failures should be easy to fix by modifying manifest files (reftest.list).
For anyone who wants to take a stab at that, the syntax is described here:
http://mxr.mozilla.org/mozilla-central/source/layout/tools/reftest/README.txt

Comment 3

[(No longer employed by Mozilla) Aaron Klotz]

The WebGL failures are covered by bug 981338.

Comment 4

[(No longer employed by Mozilla) Aaron Klotz]

(In reply to Mats Palmgren (:mats) from comment #2)
> One test, layout/reftests/bugs/759036-1.html, is an unexpected-pass.  It's
> manifest line reads:
> fuzzy-if(true,1,19) fails-if(d2d) random-if(Android&&AndroidVersion<15) ==
> 759036-1.html 759036-1-ref.html
> I guess this platform has "d2d" being true, should it?
Yes.

Comment 5

[Brad Lassey [:blassey] (use needinfo?)]

the "fails-if(d2d)" was removed by BenWa 5 weeks ago. I'm assuming this isn't an issue now?

http://hg.mozilla.org/mozilla-central/rev/1218c119a5d4

Aaron, can you retest?

Comment 6

[(No longer employed by Mozilla) Aaron Klotz]

David, I don't see test results for Windows Server 2012 on Date. Have those been disabled or something?

Comment 7

[(Away)]

Yeah, pretty much.

Jun 11 14:30:07 <philor>	dmajor: I'd sort of forgotten how date works, what I meant were the perpetually pending win64 vm tests, which won't ever run unless you get someone in releng to turn those two slaves back on, because they got shut down as a waste of time until the win64-on-hardware ones that tbpl puts in the win8 line got greened up
Jun 11 14:31:09 <dmajor>	philor: yeah, I think we only care about the hardware ones for now
Jun 11 14:31:52 <philor>	dmajor: cool, I'll keep killing the pending vm ones, which get in my way of seeing other lost pending jobs, until you tell me to stop

Comment 8

[(No longer employed by Mozilla) Aaron Klotz]

Sorry for the delay, I had to get releng to spin up the AWS VMs so I could do a push and wait for the tests to run again. The failures still exist. See https://tbpl.mozilla.org/?tree=Date&rev=7dbb7fdbbb15