780587 - Intermittent Android script-empty-01.svg | load failed: null

Status: RESOLVED INVALID

(Core :: SVG, defect)

Description

[Ed Morley [:emorley]]

Android Tegra 250 mozilla-inbound opt test reftest-3 on 2012-08-03 19:22:20 PDT for push 5d7b0289ccc7

slave: tegra-159

https://tbpl.mozilla.org/php/getParsedLog.php?id=14116798&tree=Mozilla-Inbound#error2

{
REFTEST TEST-START | http://10.250.48.218:30159/tests/layout/reftests/svg/script-empty-01.svg | 1518 / 2702 (56%)
REFTEST TEST-UNEXPECTED-FAIL | http://10.250.48.218:30159/tests/layout/reftests/svg/script-empty-01.svg | load failed: null
REFTEST INFO | Saved log: START http://10.250.48.218:30159/tests/layout/reftests/svg/script-empty-01.svg
}

(Note the scrollbars-{01,02}.svg & spellcheck-superscript-2.html failures are bug 780373 and bug 779551).

Comment 1

[Robert Longson [:longsonr]]

That's about as simple a testcase as you can get. The test checks that the script is not executed so it's basically a testcase with a lime rectangle in it comparing to another testcase with a lime rectangle in it. Something about how testcases are loaded, run and compared must be flaky on Android.

Comment 2

[Ed Morley [:emorley]]

jmaher, thoughts on comment 1? :-)

Comment 3

[Joel Maher ( :jmaher ) (UTC -8)]

there are a lot of variables on Android which could lead to confusion.

1) We run remote, so the tests are on a remote webserver and we load them remotely instead of locally (file based).
2) We are on the ARM chipset, there are always oddities here
3) We are running in Fennec.  We have a slightly reduced set of functionality and different preferences set.  

I don't know all the differences in fennec proper, but lets rope in some of the guys who might know more.

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

This bug was observed only on one or more of the six machines listed in
bug 787281 comment 11, which seem likely to have bad memory, disk, or
other hardware problem, based on the rate of failures on those machines
and the types of failures observed.

Therefore I'm marking this bug invalid, though it should be reopened if
it occurs on other (more reliable) hardware.