Using revision 19850, the pointer cursor on the image map on http://michi.stuenzi.net/ is shown everywhere, not just on the links. This doesn't occur in shipping Safari.
Probably caused by <http://trac.webkit.org/projects/webkit/changeset/19400>.
I can't get to the URL mentioned here. I'll look for other image maps.
Testing with the image map at <http://www.december.com/html/demo/imagemap.html>, everything's working fine. So I can't diagnose this without access to the broken test case.
Another test case: data:text/html,<img src="http://webkit.org/images/icon-gold.png" usemap="#map"> Some other places that check isLink() specifically exclude elements whose renderer is an image. Perhaps that's just what Document::prepareMouseEvent() should do.
The page is working again. I changed the DNS servers and forgot to set them up correctly. So please try again.
<rdar://problem/5028163>
Created attachment 13439 [details] imagemap image should never itself be the URLElement in hit test results This does not replicate Firefox behavior where if usemap attribute is empty or wrong the hand cursor is shown over entire image area. That seems more like a bug than a feature.
Created attachment 13440 [details] First attempt Hi, I am not sure HTMLImageElement should ever return true for isLink, so I am attaching the patch since it is an easy fix. Let me know if I got isLink all wrong though. Cheers, Rob.
I wasn't sure either but not setting isLink for imagemaps seemed like a risky change considering that it is used from quite many places not related to this particular problem.
I think it is safer to take Antti's narrow fix for now, and later investigate whether any of the paces calling isLink also want to include imagemaps. It's hard to tell from a quick scan of the code.
Comment on attachment 13439 [details] imagemap image should never itself be the URLElement in hit test results r=me
Comment on attachment 13440 [details] First attempt I don't think we should land this fix for now, but in the future we should figure out what parts of the code expect an imagemap to be treated as a link, and which don't.
Comment on attachment 13440 [details] First attempt (r-, but I think we should investigate this approach later, when we have time to review the affected code).
r19941