291797 - Controls using index() fail

Status: RESOLVED FIXED

(Core Graveyard :: XForms, defect)

Description

[Allan Beaufour]

If a controls tries to bind to a node using 'index()' it fails to bind with
these warnings:
WARNING: NS_ENSURE_TRUE(repeatEle) failed, file nsXFormsUtilityService.cpp, line 235
WARNING: NS_ENSURE_TRUE(NS_SUCCEEDED(rv)) failed, file XFormsFunctionCall.cpp,
line 207
WARNING: NS_ENSURE_TRUE(NS_SUCCEEDED(rv)) failed, file PredicateList.cpp, line 99
WARNING: NS_ENSURE_TRUE(NS_SUCCEEDED(rv)) failed, file LocationStep.cpp, line 231
WARNING: NS_ENSURE_TRUE(NS_SUCCEEDED(rv)) failed, file nsXPathExpression.cpp,
line 122

I think it is because we start evaluating the expression to soon. If you use
output @value, then it works, but it is also evaluated quite late. I haven't
investigated it further,,,

Comment 1

[Allan Beaufour]

Attachment: Testcase

Comment 2

[Allan Beaufour]

Attachment: Patch

Problem was the above. Controls using index() was refering to non-existant
document elements, when they were bound during initialization. I've changed
rebuild, etc. to not bind/refresh controls on initialization, and let
InitializeControls handle both bind and refresh. (This also saves an unneeded
refresh during initialization).

The patch depends on (bug 289534), and also include a leak-fix in
nsXFormsModelElement::Rebuild(): oldFormList was never deleted.

Comment 3

[Olli Pettay [:smaug][[e-mail]]]

With this patch the test case doesn't seem to work:
I see the following:

bind nodeset='value[index('repeat_id')]' = 
initially it should be == bind nodeset='value[1]' = 1

ref='value[index('repeat_id')]' = 
initially it should be == ref='value[1]' = 1

Comment 4

[Allan Beaufour]

(In reply to comment #3)
> With this patch the test case doesn't seem to work:
> I see the following:
> 
> bind nodeset='value[index('repeat_id')]' = 
> initially it should be == bind nodeset='value[1]' = 1
> 
> ref='value[index('repeat_id')]' = 
> initially it should be == ref='value[1]' = 1

It's because I'm quite stupid, and has uploaded a testcase with errors in :(

Comment 5

[Allan Beaufour]

Attachment: Testcase wo errors

Argh, I had used 'repeatid' instead of 'repeat_id' in a couple of places in the
other testcase.

Comment 6

[Doron Rosenberg (IBM)]

Comment on attachment 181865 [details] [diff] [review]
Patch

r=me

Comment 7

[Allan Beaufour]

Checked in.