Currently, WebKit would reject any compliant CSS custom cursor declaration, because it checks that there is only one value, while CSS mandates that a list of custom cursors should be followed with a named one. Even for a single value, the hot point position (as defined in CSS3) would break parsing. The attached patch lets compliant declarations parse, but doesn't implement fallback (that's bug 6001). WinIE doesn't require that a named cursor should always follow custom ones, so I don't check that there is more than one value.
Created attachment 8433 [details] proposed fix
Comment on attachment 8433 [details] proposed fix Rob is working on a real implementation that will make this hack unneeded.
The patch in bug 6002 isn't finished yet, and I think it would be wrong to introduce custom cursor support that only works with invalid declarations, so I'm reopening this bug (proposing to land this now).
Comment on attachment 8433 [details] proposed fix Looks OK, although it's missing a change log entry.
Comment on attachment 8433 [details] proposed fix Ideally we'd check that the remaining items are URIs then legal cursor names. This will allow many incorrect cursor declarations to parse too.
Hi Alexey, (In reply to comment #3) > The patch in bug 6002 isn't finished yet, and I think it would be wrong to > introduce custom cursor support that only works with invalid declarations, so > I'm reopening this bug (proposing to land this now). I got a bit busy and sidetracked, and since I am not too fond of implementing complex testcases (IIRC that was required for 6001/6002) I chose not to work on it. However IIRC it was not that far from finished, so if you feel it is important I can work on it and finish it soon. Nevertheless, any patches that improve webkit, even when they'll be replaced by other patches later, are fine by me :) Let me know on IRC how we can best approach the cursor work, if you have some time. Cheers, Rob.
Committed revision 15505. (In reply to comment #5) > Ideally we'd check that the remaining items are URIs then legal cursor names. > This will allow many incorrect cursor declarations to parse too. Agreed; I just felt that simplicity of a patch was more important in this case.