1259804 - Autophone - handling -e10s on unit tests

Status: RESOLVED INVALID

(Testing Graveyard :: Autophone, defect)

Description

[Bob Clary [:bc] (inactive)]

bug 1243083 intends to make the unit test runners default to --e10s. I ran a small set of tests with the current default which is no e10s and again with --e10s to see if there was a difference in the results.

I first did a run https://treeherder.allizom.org/#/jobs?repo=fx-team&revision=371244335c102f7e8223e522a7ed1ea98c482c3f&exclusion_profile=false&filter-tier=1&filter-tier=2&filter-tier=3&filter-searchStr=autophone without --e10s.

Then I did a run with --e10s https://treeherder.allizom.org/#/jobs?repo=fx-team&revision=d823452bea13282527bce173c1133f2f12990d7f&exclusion_profile=false&filter-tier=1&filter-tier=2&filter-tier=3&filter-searchStr=autophone and noticed several failure not in the first run.

I then re-ran the same build as used in the first run https://treeherder.allizom.org/#/jobs?repo=fx-team&revision=371244335c102f7e8223e522a7ed1ea98c482c3f&exclusion_profile=false&filter-tier=1&filter-tier=2&filter-tier=3&filter-searchStr=autophone but this time with --e10s to see if the differences were due to the additional check in and not due to the change in e10s status. The first test result is without e10s while the second is with it. The new failures with --e10s were persistent.

With --e10s I see:

C1
Android 4.0.4/samsung galaxy s3

PROCESS-CRASH | http://192.168.1.50:52796/tests/dom/base/crashtests/90613-1.html | application crashed [@ nsWindow::GeckoViewSupport::FlushIMEChanges]

Android 6.0.1/nexus 6p
PROCESS-CRASH | http://192.168.1.50:42540/tests/dom/base/crashtests/90613-1.html | application crashed [@ 0xd888d0a8]

M2
Android 4.0.4/samsung galaxy s3

234 INFO TEST-UNEXPECTED-FAIL | dom/base/test/test_bug425013.html | Missing plugin element obj1 
    runtests/<@dom/base/test/test_bug425013.html:70:5
235 INFO TEST-UNEXPECTED-FAIL | dom/base/test/test_bug425013.html | Missing plugin element embed1 
    runtests/<@dom/base/test/test_bug425013.html:71:5
236 INFO TEST-UNEXPECTED-FAIL | dom/base/test/test_bug425013.html | Missing plugin element embed2 
    runtests/<@dom/base/test/test_bug425013.html:72:5
237 INFO TEST-UNEXPECTED-FAIL | dom/base/test/test_bug425013.html | Missing plugin element obj3 
    runtests/<@dom/base/test/test_bug425013.html:73:5
238 INFO TEST-UNEXPECTED-FAIL | dom/base/test/test_bug425013.html | Wrong number of missing plugins - got +0, expected 4
    SimpleTest.is@SimpleTest/SimpleTest.js:267:5

R1
Android 4.0.4/samsung galaxy s3 and Android 6.0.1/nexus 6p

PROCESS-CRASH | http://192.168.1.50:37595/tests/layout/reftests/reftest-sanity/needs-focus.html | application crashed [@ nsWindow::GeckoViewSupport::FlushIMEChanges]

Comment 1

[Bob Clary [:bc] (inactive)]

mfinkle, snorp: Do we intend to ship e10s enabled on Android? Should I allow e10s to be enabled by default by bug 1243083 or should I force it off?

Comment 2

[Andrew Halberstadt [:ahal]]

I noticed those crashes in the emulator jobs in try too. At the time I filed bug 1243083, I didn't realize we weren't running e10s android jobs anywhere. Given that, the default for running locally shouldn't change. I'll make sure nothing changes for Android in my patch.

Bob, as long as you aren't explicitly passing --e10s in anywhere, this means there's no action required here. If you are explicitly passing it in, let me know and I'll keep it around just for Android.

Comment 3

[Bob Clary [:bc] (inactive)]

I am not currently using --e10s.