378702 - fix deferred events

Status: RESOLVED FIXED

(Core Graveyard :: XForms, defect)

Description

[aaronr]

Attachment: testcase

my patch for bug 377878 had a bug in it where I didn't correctly convert the nsisupports* back over to a nsIModelElementPrivate*.  I also previously forgot to handle the recalculate, revalidate and refresh events.  Doh!

Comment 1

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

And I reviewed bug 377878, oops.

Comment 2

[aaronr]

Attachment: patch

Comment 3

[aaronr]

(In reply to comment #1)
> And I reviewed bug 377878, oops.
> 

Nah, the bug wasn't in the patch.  The bug was the fact that I changed what went into the hashtable but I didn't change how we got the value back inside the ennumeration function.

Comment 4

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

Comment on attachment 262727 [details] [diff] [review]
patch


> PR_STATIC_CALLBACK(PLDHashOperator) DoDeferredActions(nsISupports * aModel, 
>                                                       PRUint32 aDeferred,
>                                                       void * data)
> {
>   if (aModel && aDeferred) {
>-    nsCOMPtr<nsIModelElementPrivate> model =
>-      NS_STATIC_CAST(nsIModelElementPrivate*, aModel);
>+    nsCOMPtr<nsISupports> modelSupp =
>+      NS_STATIC_CAST(nsISupports*, aModel);

aModel is already nsISupports* so no need to cast.

I think the whole problem could be fixed in some better way, but let's do this now for 0.8.
(The better way might be to use nsIModelElementPrivate as the key in the hashtable)

Comment 5

[aaronr]

(In reply to comment #4)
> (From update of attachment 262727 [details] [diff] [review])
> 
> > PR_STATIC_CALLBACK(PLDHashOperator) DoDeferredActions(nsISupports * aModel, 
> >                                                       PRUint32 aDeferred,
> >                                                       void * data)
> > {
> >   if (aModel && aDeferred) {
> >-    nsCOMPtr<nsIModelElementPrivate> model =
> >-      NS_STATIC_CAST(nsIModelElementPrivate*, aModel);
> >+    nsCOMPtr<nsISupports> modelSupp =
> >+      NS_STATIC_CAST(nsISupports*, aModel);
> 
> aModel is already nsISupports* so no need to cast.
> 

I thought we had to cast to a comptr to do a QI.  But you are right, it works w/o it.

> I think the whole problem could be fixed in some better way, but let's do this
> now for 0.8.
> (The better way might be to use nsIModelElementPrivate as the key in the
> hashtable)
> 

That's what the problem was to begin with.  QI'ing the same nsIDOMElement model to a nsIModelElementPrivate was giving back different nsIModelElementPrivate values, so looking for the values in the hashtable didn't work.

Comment 6

[aaronr]

Attachment: patch2

fixes qi nit olli found

Comment 7

[aaronr]

checked into trunk, 1.8 and 1.8.0