[webkit-reviews] review requested: [Bug 59234] CrossThreadCopier should not have a default specialization for raw pointers : [Attachment 90930] Proposed patch: second iteration

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Mon Apr 25 10:44:46 PDT 2011


Dmitry Lomov <dslomov at google.com> has asked  for review:
Bug 59234: CrossThreadCopier should not have a default specialization for raw
pointers
https://bugs.webkit.org/show_bug.cgi?id=59234

Attachment 90930: Proposed patch: second iteration
https://bugs.webkit.org/attachment.cgi?id=90930&action=review

------- Additional Comments from Dmitry Lomov <dslomov at google.com>
Code review feedback addressed.

There is one real (and quite horrible) issue that this patch uncovered - direct
your attention to changes in  IDBObjectStoreBackendImpl.cpp.
In two places, the code created a 
   RefPtr<Transaction> transactionPtr = transaction;
where transaction is Transaction*, but then passed a transaction, and not a
transactionPtr, to a scheduled task. Why this does not die a horrible death I
have no idea.
I would venture a suggestion that, besides a lax implementation of
CrossThreadCopier, a horrendous line length that pushed arguments to
scheduleTask way beyond an editor view port, was a contributing factor to this
bug.

On a subject of AllowExtendedLifetime: my suggestion would be - let us keep
AllowExtendedLifetime as an eyesore - it is something that should not be there.
ScriptExecutionContext should have a special method that only schedules tasks
on current thread.


More information about the webkit-reviews mailing list