REGRESSION (r12869): :hover subselector does not work in quirks mode

• Component: CSS
• Status: VERIFIED FIXED

The fix for bug 7218 causes div#myID:hover not to match anything in quirks mode. Need to refine it since other browsers do let it match.

Created attachment 6589 [details]
Patch to copy the tag into subselectors

I would like to fix this by having the CSS parser copy the tag into subselectors. I tried it and the only negative effect was on selectorText (and possibly on performance), and trying to fix it I realized that selectorText already has a problem with subselectors (bug 7338).

Created attachment 6684 [details]
Patch

This patch also incorporates the fix for bug 7338, but not the testcase+changelog from that bug.

[Dave Hyatt]

Comment on attachment 6684 [details]
Patch

Conceptually I'm not sure I like the idea of the tag names being propagated into these other selectors just to accommodate the :hover quirk.

Is there no way to solve this problem at the time you're testing for the quirk?

Created attachment 6698 [details]
Fix it another way

Comment on attachment 6698 [details]
Fix it another way

The quirk is more subtle than I thought. Apparently #foo:hover should match... I'll need to investigate other combinations.

Created attachment 6731 [details]
Quirks test case

Various selectors with :hover and without a tag and the browsers in which they match. ddkilzer reports that none of them match in MSIE 6. The last patch makes WebKit behave like MSIE 6.

Comment on attachment 6698 [details]
Fix it another way

According to Dave Hyatt or IRC, "we should allow hover on everything except for *", so this patch is wrong (but the right patch should be even simpler).

Created attachment 6775 [details]
Apply the *:hover and *:active quirks only when there are no sub-selectors

[Dave Hyatt]

Comment on attachment 6775 [details]
Apply the *:hover and *:active quirks only when there are no sub-selectors

r=me