54646 - :first-letter pseudo-element and :hover seem mutually exclusive

Status: RESOLVED WORKSFORME

(Core :: CSS Parsing and Computation, defect, P2)

Description

[Hixie (not reading bugmail)]

If you hover over the top of a :first-letter pseudo-element, it loses its
:first-letter style. HOW. HOW CAN THIS BE. The mind boggles.

STEPS TO REPRODUCE:
   1. http://www.webstandards.org/css/macie/pseudosdemo.html
   2. Hover over the top part of the big O in the third block (just below the
      "In the following" part of the first paragraph of Example 2).

ACTUAL RESULTS:
   1. The O shrinks.

EXPECTED RESULTS:
   1. Nothing at all! No change! Stays the same!

It's almost as if there is a :hover rule somewhere that sets all the properties
to 'inherit'. Or maybe it's the backout of joki's changes that has caused this
problem. This is just really weird though.

Comment 1

[Hixie (not reading bugmail)]

This is also visible on 
   http://www.w3.org/Style/CSS/Test/current/sec20.htm
...and is probably caused by the same problem as bug 23604.

Comment 2

[Pierre Saslawsky]

Yet another hover problem that cannot be reproduced when 
nglayout.events.showHierarchicalHover is false...

Reassigned to joki

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

You sure this isn't a dup of bug 23604 that was exposed by the new :hover code?

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

:hover is CSS2 - changing keyword

Comment 5

[Pierre Saslawsky]

You're right: this is probably a dup of bug 23604. Reassigned to attinasi.
But it means that 23604 can't be left future'd.

Comment 6

[Pierre Saslawsky]

Raising severity because it makes :first-letter unusable.

Comment 7

[Marc Attinasi]

First, I believe that bug 23604 and this are somewhat different. I investigated
23604 earlier and detailed why the bug is occurring as specific to how
WrapFramesInFirstLetterFrame is implemented vs. how normal style resolution
happens. Also, bug 23604 happens with hierarchicalHover on or off, and this one
only happens with it on.

Since this only happens with nglayout.events.showHierarchicalHover on, and it is
currently off, why is this critical?

Comment 8

[Hixie (not reading bugmail)]

no no, this happens when the pref is _off_. I have not tested it with the pref 
turned back on.

Comment 9

[Hixie (not reading bugmail)]

Taking QA per managerial policy.

Comment 10

[Christine Hoffman]

*** Bug 54927 has been marked as a duplicate of this bug. ***

Comment 11

[Marc Attinasi]

Ian, you are, of course, correct. The URL is a bit, er, tempermental in regards
to the bug. To that end I have created a simpler testcase with just a single
paragraph and a pair of rules:

  p:first-letter { color: red; }
  p:hover { color: blue; }

With these rules and a simple paragraph there is no problem, however if I put
the universal selector ('*') in front of the :hover, or just use :hover, then I
see the :first-letter getting obliterated when I hover over the text.

Attaching the testcase. I've got to keep investigating this.

Comment 12

[Marc Attinasi]

Attachment: Simple testcase that shows the bug: (put tag on :hover and bug goes away, however)

Comment 13

[Marc Attinasi]

After talking with Chris D. and looking at the bug she dup'd (bug 54927) we
discovered that this problem occurs more voraciously in Quirks mode. With a
transitional DTD, you do not even need a :hover rule to make the :first-letter
go away, whereas with a strict DTD a universal :hover is needed.

The testcase in bug 54927 is not a problem with a strict DTD, and the one I
attached to this bug IS a problem with a transitional DTD. The content model
looks the same either way, so it is likey some quirk-mode-only C++ code in
eihter the event state manager or the style system that has blown it.

Comment 14

[Marc Attinasi]

Raising Priority: this is pretty bad, since :first-letter is quite useful and
quite standard, and it won't work in many situations until this is fixed.

Ian, I think the attached testcase is pretty small, aso I removed your 'need
small testcase' note.

Comment 15

[Marc Attinasi]

I just noticed that this looks very much like bug 23604, but that was futured...

Comment 16

[Pierre Saslawsky]

*** Bug 55319 has been marked as a duplicate of this bug. ***

Comment 17

[Pierre Saslawsky]

This bug can't be reproduced in today's build (mozilla-mac-MN6 #2000-10-10-08). 
This is a Mozilla build, I haven't tried the Netscape build from ftp://sweetlou.
I guess showHierarchicalHover doesn't trigger bug 23604 anymore and we can close 
this bug after a last check with a Netscape build.

FYI, the testcase from bug 55319 pointed to the W3C CSS1 test suite:
http://www.w3.org/Style/CSS/Test/current/sec26.htm

Comment 18

[Marc Attinasi]

Works in Commercial build 2000101108 and this mornings trunk build - marking
WORKSFORME.

Any idea what fixed this?

Comment 19

[Hixie (not reading bugmail)]

I betcha that hyatt fixed this while breaking everything else with his 5693
checkin. Since he is backing out (right?) this will probably regress...

Comment 20

[Marc Attinasi]

I'm not so sure... Hyatt's checkin to nsEventStateManger that broke :hover was
made on Oct 10 13:48 (according to CVS log), but Pierre reported this fixed on
the 10-10 build, which presumably came before his change. Definitely worth
keeping an eye on when he does back it out though.

Comment 21

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Please wait until bug 5693 is resolved before verifying this bug.

Comment 22

[Hixie (not reading bugmail)]

Removing qawanted to get this off my radar.