47221 - adding datasource after load fails

Status: VERIFIED INVALID

(Core :: XPConnect, defect, P3)

Description

[Joe Hewitt]

Adding a datasource to a tree immediately after calling GetDataSource works.
However, if you wait for the datasource to have loaded, and then add it to the
tree, nothing happens.

Pretty straightforward test case coming in a segundo....

Comment 1

[Joe Hewitt]

-

Comment 2

[Joe Hewitt]

Files are at:

http://www.joehewitt.com/mozilla/testcases/47221

My stupid web host company doesn't have the xul mime type in so you can't go
straight to the file (main.xul) but you can save the files to a folder and it'll
work. Sorry about that.

Comment 3

[Chris Waterson]

Add "flex='1'" to your <treechildren> tag.

Comment 4

[Joe Hewitt]

Chris, that unfortunately does not fix the problem. It's true that I omitted 
the flex attribute, but adding it back in does not help.  Try running the demo 
and see.

Comment 5

[Chris Waterson]

I see this error using a debug build:

Document file:///d|/bugs/main.xul loaded successfully
observer: onEndLoad
************************************************************
** NOTE: This report will only be printed in DEBUG builds.**
* Call to xpconnect wrapped JSObject produced this error:  *
[Exception... "Access to XPConnect service denied. 
[nsIRDFXMLSinkObserver::onEndLoad]"  nsresult: "0x8057001e 
(NS_ERROR_XPC_JS_THREW_STRING)"  location: "<unknown>"  data: no]
************************************************************

Furthermore, if I move the files to the "chrome" directory (e.g., 
$(DIST)/bin/chrome/packages/widget-toolkit/global/content) and load from a 
"chrome:" URL, everything works fine.

jband & mstoltz might know why XPConnect doesn't want to call into JS here, 
even though UniversalXPConnect has been enabled...

Comment 6

[John Bandhauer]

No, false == "UniversalXPConnect has been enabled"

The code gets the UniversalXPConnect privilege to run the loadResultSetPost 
function.  The error happens when observer.onEndLoad() gets called sometime 
later. The privilege does not bind to the object for its lifetime. The privilege 
is stack scoped. You have to ask again in observer.onEndLoad.

Comment 7

[Joe Hewitt]

I guess this isn't a bug, then.

I am a bit surprised to find out that I have to enable XPConnect in each 
scope.  I did try once enabling in the global section of code but that didn't 
work.  Any particular reason why priviledges need to be re-enabled per scope 
rather than once per application lifetime?

Comment 8

[Joe Hewitt]

jband is correct.  enabling XPConnect after onEndLoad repairs the problem.  

Comment 9

[Mitchell Stoltz (not reading bugmail)]

Privileges are bound to scope because this limits the amount of privileged code. 
Any code for which privileges are enabled must be reviewed carefully for 
security; enabling privileges only in the function where they are needed 
minimizes the security-critical code sections.

Comment 10

[Phil Schwartau]

Verifying -