Also, I've only tested on my Mac. Might be a cross platform issue.
Thanks for the testcase.
This worksforme in my build with the fix for bug 598833, fwiw. The question is whether that's accidental...
I'm improving a part of my test suite for <input type='radio'> and I think I catch this issue by doing that...
Created attachment 533689 [details] [diff] [review] Part 1 - Improve css-invalid/input/ test suite I need tests that check the actual rendering of radio buttons. I will probably have to open some follow-ups for all similar tests.
Created attachment 533690 [details] [diff] [review] Part 2 - Actual fix. Actually, the patch is quite trivial: UpdateValueMissingValidityStateForRadio doesn't notify about |this| state change because callers have to worry about that... But one was not doing that.
BTW, there were a path for checkboxes and another for radio because the radio path was using a radio group visitor but now, UpdateAllValidityStates is going to call the correct method.
Ah, indeed. The patch in bug 598833 had added an UpdateState() call there. I guess I can remove it with this change (UpdateAllValidityStates already does UpdateState(aNotify)).
Comment on attachment 533689 [details] [diff] [review] Part 1 - Improve css-invalid/input/ test suite r=me
Pushed: http://hg.mozilla.org/mozilla-central/rev/b63ff702f39d http://hg.mozilla.org/mozilla-central/rev/fca12da605c4 Thank you Anthony for reporting this.
Follow-up for updating test suite in bug 658542.
Mozilla/5.0 (Windows NT 5.1; rv:6.0) Gecko/20100101 Firefox/6.0 Verified issue on Windows XP, Windows 7, Mac OS X 10.6 and Ubuntu using the attached test case in Comment 0. Setting resolution to VERIFIED FIXED.