870485 - Intermittent reftest failure on Windows 7 on iX machines

Status: RESOLVED FIXED

(Testing :: General, defect, P3)

Description

[Armen [:armenzg]]

Attachment: different shades of gray

https://tbpl.mozilla.org/?tree=Cedar&jobname=Windows%207%2032-bit&showall=1
https://tbpl.mozilla.org/php/getParsedLog.php?id=22776627#error0

Two pixels on two corners of a rectangle have different shades of gray (see attachment).

Comment 1

[Mats Palmgren (inactive)]

Attachment: fix

Slightly different anti-aliasing.  This should fix it.

Comment 2

[Armen [:armenzg]]

Once we have more machines we will be adding Try support and I give it a shot.
Thanks mats!

Comment 3

[Armen [:armenzg]]

I'm trying this syntax:
hg ci -m "Test - try: -b o -p win32[Windows 7 32-bit] -u reftest -t none"

The results should show up in here:
https://tbpl.mozilla.org/?tree=Try&rev=47ab47ef0b8e

Comment 4

[Armen [:armenzg]]

https://tbpl.mozilla.org/?tree=Try&rev=d9c1e1893405

Comment 5

[Phil Ringnalda (:philor)]

If it was "these slaves," then sure, just need a little more fuzzing, but it's not. It's "this one slave," t-w732-ix-016 did this on Cedar, it failed a reftest on try in some unknown and unknowable way, and then it did this exact same thing again on inbound and I filed bug 872915 on it and you disabled it. That's this machine's whole career in running reftests, three tries and three failures.

I don't actually know enough about antialiasing to know if that's reasonable, that a single machine could consistently and properly antialias differently than every other supposedly identical machine with the same graphics card. Mats?

Comment 6

[Mats Palmgren (inactive)]

Perhaps this machine thinks it's connected to a monitor of a different kind
(use different driver) or resolution than the others?

Comment 7

[Armen [:armenzg]]

I guess I can close for now. I didn't realize that it could have been a machine issue.