Closed Bug 1590604 Opened 5 years ago Closed 4 years ago

Intermittent wpt linux x64 LeakSanitizer | SUMMARY: AddressSanitizer: X byte(s) leaked in Y allocation(s) after running /service-workers/ tests

Categories

(Core :: DOM: Networking, defect, P5)

defect

Tracking

()

RESOLVED FIXED

People

(Reporter: intermittent-bug-filer, Unassigned)

References

(Blocks 1 open bug, Regression)

Details

(Keywords: intermittent-failure, regression, Whiteboard: [stockwell fixed:patch])

Filed by: dvarga [at] mozilla.com
Parsed log: https://treeherder.mozilla.org/logviewer.html#?job_id=272477626&repo=autoland
Full log: https://queue.taskcluster.net/v1/task/DLj6jqY-Qz2-k0yZ53eSAA/runs/0/artifacts/public/logs/live_backing.log


[task 2019-10-22T23:33:20.026Z] 23:33:20  WARNING - Traceback (most recent call last):
[task 2019-10-22T23:33:20.027Z] 23:33:20  WARNING -   File "/builds/worker/workspace/build/tests/web-platform/tests/tools/wptrunner/wptrunner/executors/executormarionette.py", line 627, in _run
[task 2019-10-22T23:33:20.028Z] 23:33:20  WARNING -     self.result = True, self.func(self.protocol, self.url, self.timeout)
[task 2019-10-22T23:33:20.029Z] 23:33:20  WARNING -   File "/builds/worker/workspace/build/tests/web-platform/tests/tools/wptrunner/wptrunner/executors/executormarionette.py", line 735, in do_testharness
[task 2019-10-22T23:33:20.029Z] 23:33:20  WARNING -     timeout=10*self.timeout_multiplier)
[task 2019-10-22T23:33:20.029Z] 23:33:20  WARNING -   File "/builds/worker/workspace/build/tests/web-platform/tests/tools/wptrunner/wptrunner/executors/executormarionette.py", line 216, in get_test_window
[task 2019-10-22T23:33:20.030Z] 23:33:20  WARNING -     elif handles[0] == parent and len(handles) > 2:
[task 2019-10-22T23:33:20.030Z] 23:33:20  WARNING - IndexError: list index out of range
[task 2019-10-22T23:33:20.030Z] 23:33:20  WARNING - 
[task 2019-10-22T23:33:20.031Z] 23:33:20     INFO - TEST-UNEXPECTED-ERROR | /xhr/send-sync-no-response-event-load.htm | list index out of range
[task 2019-10-22T23:33:20.031Z] 23:33:20     INFO - Traceback (most recent call last):
[task 2019-10-22T23:33:20.031Z] 23:33:20     INFO -   File "/builds/worker/workspace/build/tests/web-platform/tests/tools/wptrunner/wptrunner/executors/executormarionette.py", line 627, in _run
[task 2019-10-22T23:33:20.031Z] 23:33:20     INFO -     self.result = True, self.func(self.protocol, self.url, self.timeout)
[task 2019-10-22T23:33:20.032Z] 23:33:20     INFO -   File "/builds/worker/workspace/build/tests/web-platform/tests/tools/wptrunner/wptrunner/executors/executormarionette.py", line 735, in do_testharness
[task 2019-10-22T23:33:20.032Z] 23:33:20     INFO -     timeout=10*self.timeout_multiplier)
[task 2019-10-22T23:33:20.032Z] 23:33:20     INFO -   File "/builds/worker/workspace/build/tests/web-platform/tests/tools/wptrunner/wptrunner/executors/executormarionette.py", line 216, in get_test_window
[task 2019-10-22T23:33:20.032Z] 23:33:20     INFO -     elif handles[0] == parent and len(handles) > 2:
[task 2019-10-22T23:33:20.033Z] 23:33:20     INFO - IndexError: list index out of range
[task 2019-10-22T23:33:20.033Z] 23:33:20     INFO - TEST-INFO took 1184ms
Regressed by: 1583700
Summary: Intermittent /xhr/send-sync-no-response-event-load.htm | list index out of range → Intermittent wpt linux x64 asan /xhr/<test-name>| *test-name* | list index out of range
Keywords: regression

Modifying the summary since the real failure here are the leaks that appear when running service-worker tests.

Summary: Intermittent wpt linux x64 asan /xhr/<test-name>| *test-name* | list index out of range → Intermittent wpt linux x64 LeakSanitizer | SUMMARY: AddressSanitizer: X byte(s) leaked in Y allocation(s) after running /service-workers/ tests

Taking a quick look at a test failure, it looked like a testrunner bug. Specifically, I see "Failed to connect to Mir" ... "Gtk-WARNING **: cannot open display: :0". This is causing us to exit early, and raise an exception in the test harness.

My main guess as to how DocumentChannel could cause this is by allowing us to exceed the maximum process count & exhaust the number of clients available from the window manager.

Redirecting to :mattwoodrow who made the DocumentChannel changes, in case he has more ideas.

Flags: needinfo?(nika) → needinfo?(matt.woodrow)
Whiteboard: [stockwell disable-recommended] → [stockwell needswork]

Geoff any insight here ?

Flags: needinfo?(gbrown)
Depends on: 1595891

I notice many logs have "AttributeError: 'MarionetteProtocol' object has no attribute 'generate_test_report'", but that is present in successful runs also; I assume that is unrelated and harmless.

I don't know if the IndexError has any effect on the leaks, but I'll try to resolve that part anyway - bug 1595891.

Hopefully comment 16 will lead to a full solution.

Flags: needinfo?(gbrown)
Whiteboard: [stockwell disable-recommended] → [Comment 22][Comment 16]

Looks like bug 1595891 fixed the IndexError: list index out of range failure line here https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=276405475&repo=autoland&lineNumber=13883
The leaks are also present on green runs https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=276401583&repo=autoland&lineNumber=13805 so can we say that it is fixed?

Looks like these failures have stopped happening since the IndexError issue was fixed.

Flags: needinfo?(matt.woodrow)
Flags: needinfo?(jyavenard)
Whiteboard: [Comment 22][Comment 16][stockwell disable-recommended] → [stockwell fixed:patch]
Status: NEW → RESOLVED
Closed: 4 years ago
Flags: needinfo?(ckerschb)
Resolution: --- → FIXED
Has Regression Range: --- → yes
You need to log in before you can comment on or make changes to this bug.