Summary: | REGRESSION (r154581): Some plugin tests failing in debug bots | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Antti Koivisto <koivisto> | ||||
Component: | DOM | Assignee: | Nobody <webkit-unassigned> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | commit-queue, esprehn+autocc, kangil.han | ||||
Priority: | P2 | ||||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | Unspecified | ||||||
OS: | Unspecified | ||||||
Attachments: |
|
Description
Antti Koivisto
2013-08-26 10:04:57 PDT
Created attachment 209659 [details]
patch
Comment on attachment 209659 [details]
patch
The iterator is doing it wrong. The DOM modification assertion should not be in the iterator's destructor. Instead it should be in the iterator dereference operator. We don’t want to require people to carefully scope iterators just to time the assertion properly, and we don’t want to wait to catch the problem until long after it happened, either. The hash table iterators had the same consideration, and that's what we came up with.
(In reply to comment #2) > (From update of attachment 209659 [details]) > The iterator is doing it wrong. The DOM modification assertion should not be in the iterator's destructor. Instead it should be in the iterator dereference operator. We don’t want to require people to carefully scope iterators just to time the assertion properly, and we don’t want to wait to catch the problem until long after it happened, either. The hash table iterators had the same consideration, and that's what we came up with. DOM mutations are already tested at use time (and as you mentioned should be tested in more places). The problem here is the no-event-dispatch assertion which is an object freed in destructor. |