The page renders with an absolutely huge red section bar at the top of the page. Firefox displays it as expected. 10.4.7 also has the same issue. I'm not 100% certain, but I think its a regression with 10.4.7.
(In reply to comment #0) > I think its a regression with 10.4.7. Yes, looks fine in 10.4.6.
The script near the end of the page uses clientHeight and gets unexpected results since the stylesheet is still loading. Firefox just blocks on the stylesheet (not specifically for clientHeight). I wonder what pre-10.4.6 WebKit did.
As Mitz says, the clientHeight gets unexpected results, in this case specifically clientHeight obtains the heights of the li elements in the menu before "float: left" has been applied to them (this can be seen by removing the external style sheet, comparing heights). Also, the clientHeight obtained in each iteration of the while loop is the sum of the previous obtained clientHeights, so the height increases rapidly. I did a reduction, but it can be made clearer.
Created attachment 9364 [details] Style sheet to go with the reduction This css needs to be external.
Created attachment 9365 [details] Test case reduction Could be clearer.
(In reply to comment #1) > (In reply to comment #0) > > I think its a regression with 10.4.7. > > Yes, looks fine in 10.4.6. > Going over the changes between 10.4.6 and 10.4.7 I can't spot anything that could explain a change in the general behavior, but only a fix related to height and max-height. Could you please check in 10.4.6 whether the script executes while the stylesheet is still loading, and if it does, what clientHeight (and other layout properties) return at that point?
This reduction fails on 10.4.6 too, so it probably illustrates a somewhat different aspect of the problem. Since I'm likely the only one with 10.4.6 still around, I'll try to make a reduction that illustrates the regression. I think that the problem with stylesheet loading not blocking JS is a duplicate.
(In reply to comment #7) > This reduction fails on 10.4.6 too, so it probably illustrates a somewhat > different aspect of the problem. Since I'm likely the only one with 10.4.6 > still around, I'll try to make a reduction that illustrates the regression. I have a feeling that would be fast/css/max-height-none.html . I think fixing rdar://problem/4335038 just exposed another instance of the issue of not blocking on a stylesheet. > I think that the problem with stylesheet loading not blocking JS is a > duplicate.
(In reply to comment #8) > I think fixing > rdar://problem/4335038 just exposed another instance of the issue of not > blocking on a stylesheet. That's wrong. Must have been some other fix, since that one is new and this bug's URL is broken in the 2005-10-01 nightly.
Oops, now it has started to happen on stock 10.4.6, too :-/. I saw it render correctly 3-4 times before (and always fail with a nightly), but it must have all been random. Removing regression keywords.
Duplicate of bug 8852?
Created attachment 13912 [details] Test case reduction Corrected OpenDarwin bugzilla reference.
Safari, Chrome, and Firefox show the same behavior on the reduced test case.