Prevent WPT tests from potentially affecting other tests
Categories
(Core :: DOM: Notifications, task)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox128 | --- | fixed |
People
(Reporter: saschanaz, Assigned: saschanaz)
References
Details
Attachments
(3 files)
Running ./mach wpt /notifications/constructor-basic.https.html /notifications/constructor-invalid.https.html for example causes Notification::CloseInternal being called for the notifications opened by the first one while running the second one. This technically shouldn't affect the second test itself, but we shouldn't run any task after the test finish to prevent confusing intermittent results.
| Assignee | ||
Comment 1•2 years ago
|
||
What's weird is that we shouldn't even call CloseInternal as constructor-basic has no permission to open anything 🤔
| Assignee | ||
Updated•2 years ago
|
| Assignee | ||
Comment 2•2 years ago
|
||
| Assignee | ||
Comment 3•2 years ago
|
||
Fixing one test to set permission properly and another to not trigger external network request.
Comment 6•2 years ago
|
||
Backed out for causing mozilla::PermissionManager::CommonTestPermissionInternal related wpt crashes.
- Backout link
- Push with failures
- Failure Log
- Failure line: PROCESS-CRASH | MOZ_ASSERT(PermissionAvailable(prin, aType)) [@ mozilla::PermissionManager::CommonTestPermissionInternal] | /notifications/permission.html
| Assignee | ||
Comment 7•2 years ago
|
||
But it was clean in https://treeherder.mozilla.org/jobs?repo=try&revision=4502ed827177847cc26dabd18a955c273becaede 😞
| Assignee | ||
Comment 8•2 years ago
|
||
So it's always permission.html, at least that's clear. For now I'll just mark it as CRASH as this is blocking other test changes and I think this crash has been there for a while, if not forever.
Comment 10•2 years ago
|
||
Hey @saschanaz, are you planning to push these changes again soon? I just have the stack on bug 1524074 which I rebased on these changes, and now they are ready to go. If you don't have time at the moment to finish, I can revert the update and land my stack, and then you will have to rebase on my changes. Let me know, what works for you.
| Assignee | ||
Comment 11•2 years ago
|
||
I think I'll investigate a bit more, so feel free to land first 👍 (I locally modified the test and moved the current test to crashtest, and then the crashtest does not crash anymore, so 🤷♀️)
| Assignee | ||
Comment 12•2 years ago
|
||
This is sad but at least we'll have a test that fails so that GeckoView people can take a look.
Removing OK/TIMEOUT as I think that's just bug 1798871.
Comment 13•2 years ago
|
||
Comment 14•2 years ago
|
||
| bugherder | ||
https://hg.mozilla.org/mozilla-central/rev/51eecbe291e0
https://hg.mozilla.org/mozilla-central/rev/5ab2cff5dfce
https://hg.mozilla.org/mozilla-central/rev/03da35eb0a10
Description
•