Bug 32501 - AX: there's an infinite loop going on for some webpages - facebook.com
Summary: AX: there's an infinite loop going on for some webpages - facebook.com
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Accessibility (show other bugs)
Version: 528+ (Nightly build)
Hardware: PC OS X 10.5
: P2 Normal
Assignee: chris fleizach
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2009-12-13 16:22 PST by chris fleizach
Modified: 2009-12-14 08:25 PST (History)
2 users (show)

See Also:


Attachments
patch (99.93 KB, patch)
2009-12-13 16:48 PST, chris fleizach
abarth: review-
Details | Formatted Diff | Diff
patch (3.84 KB, patch)
2009-12-13 18:12 PST, chris fleizach
no flags Details | Formatted Diff | Diff
patch (9.17 KB, patch)
2009-12-13 23:18 PST, chris fleizach
adele: review+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description chris fleizach 2009-12-13 16:22:39 PST
probably not infinite, but way too slow
Comment 1 chris fleizach 2009-12-13 16:48:05 PST
Created attachment 44769 [details]
patch

the problem is that we can't call parentObjectUnignored while in accessibilityIsIgnored... it leads to a lot of recursiveness that might not ever end...

it was being done when determining if an object was within a ARIA tree. that check should
1) be moved to all platforms
2) not call parentObjectUnignored()
Comment 2 WebKit Review Bot 2009-12-13 16:48:41 PST
style-queue ran check-webkit-style on attachment 44769 [details] without any errors.
Comment 3 Adam Barth 2009-12-13 16:52:33 PST
Comment on attachment 44769 [details]
patch

That test case is huge.  Is it possible to test the issue with a reduced test case?

More problematically, that HTML looks to be a derivative work of Facebook.  I suspect we don't have the right to re-license it under the BSD license.
Comment 4 chris fleizach 2009-12-13 18:10:13 PST
the bug looks like its a result of the huge number of elements on the page, not in any particular snippet. i can only resubmit without a layout test
Comment 5 chris fleizach 2009-12-13 18:12:01 PST
Created attachment 44773 [details]
patch

this patch omits the layout test. the slow down captured by this bug is a result of pages that have deeply nested hierarchy and a lot of it. i can't simulated a page without copying another webpage
Comment 6 WebKit Review Bot 2009-12-13 18:15:50 PST
style-queue ran check-webkit-style on attachment 44773 [details] without any errors.
Comment 7 chris fleizach 2009-12-13 22:41:08 PST
Comment on attachment 44773 [details]
patch

need to see if i can create a test
Comment 8 chris fleizach 2009-12-13 23:18:56 PST
Created attachment 44774 [details]
patch

made a test that gets the behavior. we just needed a lot of elements inside of each other
Comment 9 WebKit Review Bot 2009-12-13 23:19:59 PST
style-queue ran check-webkit-style on attachment 44774 [details] without any errors.
Comment 10 chris fleizach 2009-12-14 08:25:43 PST
http://trac.webkit.org/changeset/52095