[Webkit-unassigned] [Bug 153038] New: Modern IDB: A few cursor tests are flaky because JS wrappers are GC'ed

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=153038

            Bug ID: 153038
           Summary: Modern IDB: A few cursor tests are flaky because JS
                    wrappers are GC'ed
    Classification: Unclassified
           Product: WebKit
           Version: Safari 9
          Hardware: Unspecified
                OS: Unspecified
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: WebCore Misc.
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: beidson at apple.com

Modern IDB: A few cursor tests are flaky because JS wrappers are GC'ed

storage/indexeddb/cursor-update.html
storage/indexeddb/mutating-cursor.html

https://build.webkit.org/results/Apple%20Yosemite%20Release%20WK1%20(Tests)/r194903%20(11144)/results.html
https://build.webkit.org/results/Apple%20Yosemite%20Release%20WK1%20(Tests)/r194901%20(11142)/results.html

You can see on the flakiness dashboard that on debug bots they are crashes:
https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#tests=storage%2Findexeddb%2Fmutating-cursor.html

But really ASSERTs.

I verified running these tests repeatedly with some logging that the IDBRequest answered "false" to hasPendingActivity(), leading to its JS wrapper to get GC'ed, but then an event was delivered.

The ASSERTs in debug builds are just that: 
inline JSC::JSObject* JSEventListener::jsFunction(ScriptExecutionContext* scriptExecutionContext) const
...
    ASSERT(!m_isolatedWorld->isNormal() || m_wrapper || !m_jsFunction);

IDBRequests for cursors operations can be reused and must maintain a "true" answer to this question while a cursor operation is outstanding.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20160112/267d904a/attachment.html>