342490 - window.navigator isn't really readonly

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Blake Kaplan (:mrbkap) (inactive)]

In nsIDOMWindowInternal.idl, we have:
http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/dom/public/idl/base/nsIDOMWindowInternal.idl&rev=1.32&root=/cvsroot&mark=54#54

However, XPConnect always asks the scriptable helpers to resolve a property before falling back onto asking the set if the property exists. As a result, we fall into this code in nsDOMClassInfo.cpp:
http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/dom/src/base/nsDOMClassInfo.cpp&rev=1.393&root=/cvsroot&mark=6289-6291#6276

and since we resolve it there, we never ask the IDL about the property, and we lose. I'm not sure why we have the DOMClassInfo code there at all (why don't we just resolve the navigator property in from the IDL?) but if there is a need for it, then I think we should use JSPROP_READONLY and JSPROP_PERMANENT for it.

Comment 1

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: Fix for all branches

This patch is good for all branches. It fixes this bug by making the 'navigator' property permanent and readonly.

Comment 2

[Johnny Stenback (:jst)]

Comment on attachment 226849 [details] [diff] [review]
Fix for all branches

r+sr=jst

Comment 3

[Blake Kaplan (:mrbkap) (inactive)]

Fix checked into trunk.

Comment 4

[Robert Sayre]

Attachment: refresh to see window.navigator writable

Comment 5

[Robert Sayre]

window.navigator is still writable if you refresh attachment 244036 [details].

Comment 6

[Robert Sayre]

Attachment: set window.navigator = document;

Comment 7

[Robert Sayre]

Attachment: this ones shows it's not a printing bug

Comment 8

[Johnny Stenback (:jst)]

Attachment: Additional change needed...

Comment 9

[Blake Kaplan (:mrbkap) (inactive)]

Comment on attachment 244254 [details] [diff] [review]
Additional change needed...

Ah, drat.

Comment 10

[Johnny Stenback (:jst)]

Fixed on the trunk.