280708 - [FIX]hang/freeze when hovering over input (using input:hover,td:hover {display:table-row-group}

Status: VERIFIED FIXED

(Core :: Layout: Tables, defect, P1)

Description

[Martijn Wargers (dead)]

See upcoming testcase. 
Current trunk builds hang/freeze when hovering over the <input> in the testcase.
This happens because I Used 
input:hover{display:table-row-group;}
td:hover{display:table-row-group;} 
as css rules.

Comment 1

[Martijn Wargers (dead)]

Attachment: Testcase

Comment 2

[Bernd]

Martijn, as usual bugs like this need to get prioritized. The main criteria for
me is regression. Can you try this with FF 1.0 (class:somehow old, should be
fixed for 1.1) and even Seamonkey 1.6 (class: ancient, we need to fix this but
1.1 is not a milestone)

Comment 3

[Martijn Wargers (dead)]

Ah, ok. I'm testing (semi-)randomly, that's why I wasn't testing it in older
builds, but I'll do that from now on.

It doesn't freeze/hang with:
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b) Gecko/20050124 Firefox/1.0+
But it does freeze hang with:
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b) Gecko/20050125 Firefox/1.0+

Maybe a regression from fixing bug 269566?

Comment 4

[Boris Zbarsky [:bzbarsky]]

Yeah, probably... I'll look into this one.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Attachment: Patch

This is effectively the same as the patch for bug 275746 (and backs that patch
out, really), but I've pushed the code down into CreateAnonymousFrames, since
all callers of the method would need to do this....  In particular, <input>
creates anonymous frames for the editor inside it, so needs this.

Comment 6

[Bernd]

Boris, is this really correct given the fact that even ConstructTableFrame calls
CreateAnonymousFrames?

Comment 7

[Boris Zbarsky [:bzbarsky]]

I think it's correct, yes.  Note that the very first line in
CreateAnonymousFrames (the version that ConstructTableFrame calls, not the one I
changed; the latter is called by the former) bails out if the node is not a root
(which is a table never claims to be) and if the tagname isn't in a specific list.

Also note that I'm only doing this pseudo thing if the tag is one we want to
craete anonymous content for (so the inner method got called) _and_ the parent
frame QIs to nsIAnonymousContentCreator (which table frames do not) _and_ it
creates anonymous content.

This isn't the "totally right" fix, since anonymous content with table display
types should really share pseudo-frames with non-anonymous kids with table
display types.  But that's something that would be easier to fix with a general
pseudo-frames revamp...

Comment 8

[Boris Zbarsky [:bzbarsky]]

Just for reference, the parts of the diff that deal with iterators and
ProcessAttachedQueue were not supposed to be there... I'll remove them before
checking in.

Comment 9

[Boris Zbarsky [:bzbarsky]]

Fixed.

Comment 10

[Ger Teunis]

Upon hovering over the test case, the text box moves off to the right. Is this
expected behaviour?

Comment 11

[Martijn Wargers (dead)]

Ger, that's a known bug (don't know the bug number), but unrelated to this bug.
When this bug is fixed (which it should in 20050204 trunk build), it should
indeed show that behavior you described (and not hang/freeze).

Comment 12

[Martijn Wargers (dead)]

Verified using 20050204 trunk build.

Comment 13

[Boris Zbarsky [:bzbarsky]]

Checked in a crashtest and reftest for this.

Comment 14

[Robert O'Callahan (:roc) (email my personal email if necessary)]

The reftest 280708-1a.html actually failed on all Tinderbox platforms, so I marked it as expected-fail and filed bug 473824 for it.