| Summary: | IndexedDB: De-flake open-during-transaction layout test | ||||||
|---|---|---|---|---|---|---|---|
| Product: | WebKit | Reporter: | Joshua Bell <jsbell> | ||||
| Component: | New Bugs | Assignee: | Joshua Bell <jsbell> | ||||
| Status: | RESOLVED FIXED | ||||||
| Severity: | Normal | CC: | alecflett, dgrogan, tony, webkit.review.bot | ||||
| Priority: | P2 | ||||||
| Version: | 528+ (Nightly build) | ||||||
| Hardware: | Unspecified | ||||||
| OS: | Unspecified | ||||||
| Attachments: |
|
||||||
|
Description
Joshua Bell
2013-02-06 11:05:48 PST
Created attachment 186879 [details]
Patch
dgrogan@, alecflett@ - can you take a look? I saw this flake when doing repeated runs of the storage/indexeddb tests. I was actually surprised to see this flake under DRT (single-process) as opposed to content_shell (multi-process), but I believe it's due to the timing of the open() calls occurring in event callbacks vs. the transaction timer. (I also moved my /tmp dir to a ramdisk so the opens() take less wall-clock time on my machine.) I don't know if it's worth digging in further. The test was first added in wkrev.com/98749 to act as a canary for crashes around with metadata snapshotting. Since then that whole mechanism has changed and there are dedicated tests and setVersion() has been removed. So I think the remaining purpose of this test is merely "make sure open()-with-no-version isn't blocked on a running transaction". This seems trivial but weird, so feedback appreciated. lgtm lgtm2 tony@ - when convenient, r? cq? Comment on attachment 186879 [details] Patch Clearing flags on attachment: 186879 Committed r142324: <http://trac.webkit.org/changeset/142324> All reviewed patches have been landed. Closing bug. |