86755 - anonymous outlinerbody's database property is null in containing XBL binding's constructor.

Status: RESOLVED FIXED

(Core Graveyard :: RDF, defect, P3)

Description

[Eric Murphy]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9.1+)
Gecko/20010618
BuildID:    2001061804

The summary says it all.

Reproducible: Always
Steps to Reproduce:
Use the testcase, and view the dump.

Comment 1

[Eric Murphy]

Attachment: roster.xml (testcase)

Comment 2

[Eric Murphy]

Also, using the same approach, except with the outliner being in the
<children/>, I get the same result.

Comment 3

[Chris Waterson]

Could be pilot error, but I'm having a hard time getting outliner to work _at
all_ as an XBL-ified widget. See bug 87860.

Comment 4

[Ben Goodger (use ben at mozilla dot org for email)]

My suspicion here is that you're accessing .database on outlinerbody before it
exists. I created an example that accesses .database as the result of a button
click, and it was non-null. 

Seems like something is fishy here, but it's probably possible to use any number
of hacks to work around the issue. 

Resummarizing to reflect this. 

Comment 5

[Ben Goodger (use ben at mozilla dot org for email)]

Yep, this is confusing. This is my hypothesis. I may be wrong.

The problem appears to be because no frame has been created for the outlinerbody
element. 

Description:

1) the outliner builder calls |SetView| on the outliner body frame, 

2) this in turn calls |SetOutliner| on the outliner builder passing its outliner
box object. 

3) |SetOutliner| on the outliner builder calls |GetOutlinerBody| on the outliner
box object passed to it. 

4) In the <xbl:constructor/> of a binding containing an anonymous outliner,
|GetOutlinerBody| returns null, as it is unable to get the frame for the
outlinerbody element, as that has not yet been created. 


Well that's my debugging fun for the evening! 

Comment 6

[Eric Murphy]

I found a good hack for this. Move the code from the binding's constructor to a
method. Then in the document the binding is used in do

    window onload="thebinding.theMethod()"

and that will execute the original constructor code after the outliner has
loaded its database.

Comment 7

[Chris Waterson]

Yeah, what we need to do here is make nsXULTemplateBuilder (the super-class for
nsXULOutlinerBuilder) inherit from nsIRDFContentModelBuilder. Then we'll just
call nsIRDFContentModelBuilder::SetRootContent when we're in
nsXULDocument::CheckTemplateBuilder, so that the |database| property gets
initialized regardless of whether there's a frame or not.

Comment 8

[Chris Waterson]

Attachment: fix

The problem was that we weren't adding the database property in the
nsXULOutlinerBuilder until the frame was created and the outliner box object
was being set. I fixed this by getting rid of nsIRDFContentModelBuilder (which
wasn't really necessary), moving those methods to nsIXULTemplateBuilder, and
having the initialization code in nsXULDocument::CheckTemplateBuilder simply
call the `Init' method (formerly `SetRootContent') regardless of whether it was
initializing a content builder or an outliner builder.

I'll attach a test case that can be used to verify the bug.

Comment 9

[Chris Waterson]

Attachment: xul file, save as `86755.xul'

Comment 10

[Chris Waterson]

Attachment: CSS file, save as `86755.css'

Comment 11

[Chris Waterson]

Attachment: XBL file, save as `86755.xml'

Save these files, and JAR them up into a chrome JAR somewhere. Then you'll be
able to run the test and see the `dump' statement in the XBL file spew
something other than `null'.

Comment 12

[Chris Waterson]

Attachment: complete patch

Sorry, attachment 66068 [details] [diff] [review] was incomplete (missing diffs from
mozilla/content/build). This patch is complete.

Comment 13

[Jan Varga [:janv]]

Nice work! Now it looks even more consistent.
This patch also fixes some problems I pointed out in bug 120253.

+            // Force an <outlinerchildren> to be created if one isn't
+            // there already.
+            // XXXwaterson why do we have to do this?

I think this is only way how to create an outliner body for the rdfliner.


r=varga

Comment 14

[Chris Waterson]

Gotcha. I'll fix the comment to say just that.

Comment 15

[David Hyatt]

sr=hyatt

Comment 16

[Chris Waterson]

Fix checked in.

Comment 17

[benc]

tever is not RDF QA anymore