Bug 21487
| Summary: | JSCustomSQLTransactionCallback and associated objects leak when the storage tests are run | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | mitz |
| Component: | WebCore Misc. | Assignee: | Dumitru Daniliuc <dumi> |
| Status: | RESOLVED WORKSFORME | ||
| Severity: | Normal | CC: | beidson, dumi, eric |
| Priority: | P1 | Keywords: | InRadar, LayoutTestFailure |
| Version: | 528+ (Nightly build) | ||
| Hardware: | Mac | ||
| OS: | OS X 10.5 | ||
mitz
./WebKitBuild/Debug/DumpRenderTree LayoutTests/storage/*.html reports the following:
LEAK: 19 WebCoreNode
LEAK: 1 JSCustomSQLTransactionCallback
LEAK: 1 Frame
LEAK: 44 JSC::Node
The first test in the bunch, close-during-stress-test.html, is essential for the leak to happen, and running it (or even a reduced version of it) alone reproduces the leak, but running it followed by a page that waits for a while before DRT quits makes the leak go away.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
mitz
<rdar://problem/6279671>
Eric Seidel (no email)
This may be related to the crash (bug 30172) that we're seeing with storage/close-during-stress-test.html
Dumitru Daniliuc
Just sync'ed to r59499 and ran WebKitBuild/Debug/DumpRenderTree LayoutTests/storage/*.html:
Leak: 89 WebCoreNode
Leak: 616 Structure
Running only close-during-stress-test.html doesn't report any leaks.
All DB bindings were refactored quite a bit in the last half a year. One of those refactorings must've fixed this problem too. Please reopen if you this leak again.