Open Bug 1544637 Opened 6 months ago Updated 5 months ago

WPT on Geckoview frequently crashes in various feature-policy tests

Categories

(Core :: DOM: Security, defect, P3)

defect

Tracking

()

Tracking Status
firefox68 --- affected

People

(Reporter: KWierso, Unassigned)

References

(Blocks 1 open bug)

Details

(Whiteboard: [geckoview:p2][domsecurity-backlog2])

I'm trying to get web-platform-tests running against Geckoview's test runner activity, and I'm seeing some frequent (but not necessarily permanent) crashes during various feature-policy tests. I can try to mark them as expected crashes to green up the test runs, but figured I should file a bug to track actually fixing the crashes.

The crashing tests:
/feature-policy/payment-allowed-by-feature-policy.https.sub.html
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=239765432&repo=try&lineNumber=3105
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=239860275&repo=try&lineNumber=3103
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=239860307&repo=try&lineNumber=3103
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=239860353&repo=try&lineNumber=3099

/feature-policy/payment-default-feature-policy.https.sub.html
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=239860275&repo=try&lineNumber=3129
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=239860353&repo=try&lineNumber=3125

/feature-policy/payment-disabled-by-feature-policy.https.sub.html
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=239765432&repo=try&lineNumber=3135

(I also see a number of crashes in various payment-request tests, so maybe this crash is related to the payment APIs themselves, and not the feature-policy parts?)

Sam, Baku, any thoughts on the failing tests?

Flags: needinfo?(sfoster)
Flags: needinfo?(amarchesini)

(In reply to Christoph Kerschbaumer [:ckerschb] from comment #1)

Sam, Baku, any thoughts on the failing tests?

None I'm afraid. The web-platform tests were mostly Baku and :marcosc I think?

Flags: needinfo?(sfoster) → needinfo?(mcaceres)

Is it possible to have a stack trace? It would be extremely useful to identity which component crashes and why. Thanks.

Flags: needinfo?(amarchesini) → needinfo?(wkocher)

Working on finding out what it'll take to get usable stack traces in Bug 1544621. I don't think any of these bugs I filed are actionable until that happens, so no rush to deal with these. I haven't even fully turned the geckoview wpt tests on in production yet.

Flags: needinfo?(wkocher)
Flags: needinfo?(mcaceres)
Priority: -- → P3
Whiteboard: [domsecurity-backlog2]
Whiteboard: [domsecurity-backlog2] → [geckoview]
Whiteboard: [geckoview] → [geckoview][domsecurity-backlog2]

Here's a run with crash stacks: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=243015925&repo=try&lineNumber=3091

Unsure if these are actual crashes or if there's an issue with the test harness that forces the app to crash out of a bad situation.

Flags: needinfo?(mcaceres)

Baku might have hunches on those stacks. (link comment 5)

Flags: needinfo?(amarchesini)
Flags: needinfo?(mcaceres)
See Also: → 1495301

I'm going to disable the Payments tests as we are no longer working on Payment Request API.

Depends on: 1549241
Flags: needinfo?(amarchesini)
Depends on: 1550412

(In reply to Wes Kocher (:KWierso) from comment #5)

Here's a run with crash stacks: https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=243015925&repo=try&lineNumber=3091

Unsure if these are actual crashes or if there's an issue with the test harness that forces the app to crash out of a bad situation.

Unfortunately, the logs from that Try run are no longer available. The logviewer summary says:

01:34:35 WARNING - InvalidArgumentException: Unknown pointerType: [object String] "touch"
01:34:35 WARNING - InvalidArgumentException: Unknown pointerType: [object String] "touch"
01:35:13 INFO - PROCESS-CRASH | mozrunner-startup | application crashed [@ libc.so + 0x8dd27]
01:35:18 INFO - PROCESS-CRASH | mozrunner-startup | application crashed [@ google_breakpad::ExceptionHandler::WriteMinidump()]

I don't know if the libc.so PROCESS-CRASH is a real crash or just how the mozrunner-startup program is reporting the test failure.

Whiteboard: [geckoview][domsecurity-backlog2] → [geckoview:p2][domsecurity-backlog2]
You need to log in before you can comment on or make changes to this bug.