[Webkit-unassigned] [Bug 83663] REGRESSION (r104445): Style is not recomputed on serenaandlily.com

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=83663





--- Comment #3 from Warren <wherethebandsr at yahoo.com>  2012-04-14 15:56:12 PST ---
Created an attachment (id=137220)
 --> (https://bugs.webkit.org/attachment.cgi?id=137220&action=review)
5 test cases (each being a single .html file) that illustrate the exact nature of the bug

As reported at http://crbug.com/122663:

The issue boils down to css identifiers with a value that uses CamelCase!

I've attached 5 (relatively) minimal tests that illustrate this behavior.
They only differ by a few characters.. so the easiest way to see what has been tweaked is by looking at a diff, but I'll summarize.

test #1 (pass):
  the css id="div_lowercase"
test #2 (fail):
  the css id="div_CamelCase"
test #3 (fail):
  the css id="div_CamelCase"
  addendum:
  * test #2 wasn't concerned with bumping up the css width of #div_02
    the first css block sets its width = 729px
    which is why when #div_CamelCase is (incorrectly) still visible,
    the 2 floats can still fit side by side without vertically wrapping.
  * test #3 illustrates that we can add a rule
    in the same <style> block that contains the rule to hide #div_CamelCase,
    which is being ignored,
    to increase the width of #div_02.
    this is an important test because all selectors in the 1st style tag
    specify a more precise class hierarchy and (consequently) take precedence.
    both of the rules in the 2nd style block use "!important" to over rule.
    this test shows that this is not the issue,
    because it works for one but not the other (ie: the one with CamelCase).
test #4 (pass):
  the css id="div_CamelCase"
  addendum:
  * test #4 is fun because:
    instead of using "!important" in the 2nd style block to over rule the 1st,
    #div_02 uses the same class hierarchy to take precedence (since it occurs later).
    what's really weird here,
    is that even though the rule to hide #div_CamelCase has NOT been changed,
    and the rule that has been change is effectively equivalent to test #3,
    #div_CamelCase is NOW HIDDEN.
test #5 (pass):
  the css id="div_CamelCase"
  addendum:
  * test #5 is a slight modification on what was shown in #4
  * the width of #div_02 is ignored (as it was in #2)
  * the selector in the 2nd style block for #div_CamelCase is made more explicit,
    and suddenly it is obeyed!

so.. long story short:
  * somewhere in WebKit..
    case sensitivity has been introduced in css selector parsing.
  * when a css id is impacted by this bug,
    there are some weird ways to work around it.

hope this helps..

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.