Closed Bug 1188052 Opened 4 years ago Closed 3 years ago

Frequent OSX 10.6 debug dom/plugins mochitest hangs leading to harness timeouts

Categories

(Core :: Plug-ins, defect)

x86_64
macOS
defect
Not set

Tracking

()

RESOLVED INCOMPLETE
Tracking Status
firefox42 --- affected

People

(Reporter: RyanVM, Unassigned)

References

Details

Not a new problem exactly, but it seems to have gotten worse lately. Given that it happens only on OSX 10.6 debug, aklotz has already told me that he wouldn't be heartbroken if we just skipped these tests on that combo.
Keywords: leave-open
test_beaconCookies.html started permafailing after dom/plugins was skipped, so I'm skipping it on 10.6 also and hoping the carnage stops there.
https://hg.mozilla.org/mozilla-central/rev/19c842fad15c
I had to back out all of this because m(oth) was still permafailing and I'm tired of fighting it tonight.
Flags: needinfo?(ryanvm)
Which leaves us with the exciting conundrum of what to do next. We have a test chunk that isn't meeting visibility standards (>5% failure rate) and disabling our way to victory has proven to be horrendously ill-fated.

Aaron, any suggestions?
Flags: needinfo?(ryanvm) → needinfo?(aklotz)
Do we have any data as to when these timeouts spiked?
Flags: needinfo?(aklotz) → needinfo?(ryanvm)
Kyle, is that something ActiveData could help with?
Flags: needinfo?(ryanvm) → needinfo?(klahnakoski)
Let me look into the the duration for these tests/jobs/chunks and I will get back to you.  We do not have visibility to the timeouts, but maybe we can see a pattern in the holes.
How far back do we expect this problem to be?  I looked back to June 15 and I see nothing that would indicate a change in behavior, nor any creeping change that may hit a timeout limit.  I am still looking at some smaller anomalies.
Flags: needinfo?(klahnakoski)
The problem is that these issues have always been around to some extent (seriously, look at how old most of the deps are), so it's hard to get a feel for when things got markedly worse.
Going through the various deps, I'd say to maybe start looking around mid-July, but that's a bit hand-wavey.
I broke down average run times for all tests [1].  They all seem quite stable.  

* dom/plugins/test/mochitest/test_bug1165981.html was added around July 21st, 
  but its runtime is very short ad makes no overall impact
* There were failures and slowdowns around July 5 and July 6, but normal afterward

Am I looking at the right thing (example [2])?

* All tests atarting with "dom/plugins"
* in the "mochitest" suite
* using the "debug" build
* and run on "yosemite"

[1] http://activedata.allizom.org/tools/query.html#query_id=dw285BlR
[2] http://activedata.allizom.org/tools/query.html#query_id=u1c1+Hi2
Blocks: 1192246
Depends on: 648525
Resolving old bugs which are likely not relevant any more, since NPAPI plugins are deprecated.
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → INCOMPLETE
Removing leave-open keyword from resolved bugs, per :sylvestre.
Keywords: leave-open
You need to log in before you can comment on or make changes to this bug.