[jQuery] $(document).ready still fires too early in svn trunk (rev 4000)

[moe]

hi,

i spent the last half hour reading up on this and it
seems that this bug was supposed to be fixed. at least
the only related and still open ticket that i can find
in trac refers to safari - which i'm not testing.

well, here it strikes again.

i can unfortunately not provide a complete test-case as
i haven't managed to reproduce it from scratch and the
page where it happens is private (corporate foobar).

well, here's the meat that i can provide for now:

* works well in firefox

* fires too early in IE7 and opera

* seems to only affect reasonable complex pages
  where load/render times come into play

* happens in conjunction with roundcorners-canvas from
  http://code.google.com/p/jquery-roundcorners-canvas/

the code that triggers the problem:

$(function() {
   $('#bigbox').corner("round 5px");
}

this works fine in firefox but blows up in opera
(every time) and in IE7 (roughly every second time).
wrt opera you may take "blows up" quite literally
as it annihilates the page layout to the point of
unreadability; page gets overwidth, floats jump
to just about anywhere. seems like roundcorners-
canvas does some funny things when the DOM 
(or something else?) is not complete. i
even had opera crash multiple times.
it's not as bad in IE7, there it just fails
silently (no rounding) without further impact.


regardless i can make it work in all browsers (100%)
using the following workaround:

$(function() {
   setTimeout(
     function() {
       $('#bigbox').corner("round 5px");
     }, 500
   );
}

i'm positive that this must be a timing issue because
it also strikes when i remove all other javascript
and css, except the includes of jquery and
roundcorners-canvas.

it doesn't seem to matter where i include the above code snippets.
they behave the same when included inline at the top or bottom
(before </body>) of the page and when included from an external
js file. i also get identical behaviour when i omit the
$(function())-wrapper, except that the variant without
the timer will then also fail in firefox.

so, am i missing something obvious here (ouch!) or
has our ol'friend the timing bug returned?


regards, moe