11454 - xpidl doesn't respect [noscript] for attributes

Status: RESOLVED FIXED

(Core :: XPCOM, defect, P3)

Description

[Mike McCabe]

Given idl with a line such as

    [noscript] attribute string dummy;

xpidl won't generate 'hidden' bits for the accessor methods in the .xpt file,
and XPConnect will happily access the attribute.  This turned up because
somebody has an 'attribute voidStar foo' in their .idl file, and we'd like to
apply the same requirement that it be marked [noscript] that interface methods
will soon have.  (When I check in...)

/mailnews/base/search/public/nsIMsgSearchSession.idl has
     readonly attribute voidStar searchParam;

Comment 1

[Mike McCabe]

Marking as assigned.

A logical extension of the [noscript] error recently checked in.

Comment 2

[Mike McCabe]

*** Bug 12187 has been marked as a duplicate of this bug. ***

Comment 3

[Mike McCabe]

I need to fix nsISupportsPrimitives.idl when I fix this.

Comment 4

[Mike McCabe]

Fix checked in.  Still in question is whether xpidl_typelib generates the 'don't
script me' bits, which I'm not sure about.  Leaving open for now.

Comment 5

[John Bandhauer]

So far so good. xpt_dump shows that the 'hidden' bit is not being set. This is 
worth fixing. It looks like typelib_attr_accessor is the place to set it. 
typelib_op_dcl has a fine example of how. You just have to get the 
"noscript" prop from a IDL_ATTR_DCL instead of a IDL_OP_DCL. and set 
XPT_MD_HIDDEN or not. We know that attributes can't be XPT_MD_NOTXPCOM.

You've already done the hard part of fixing up the idl files in the tree.

Comment 6

[Mike McCabe]

Attachment: diff to xpidl_typelib.c to mark [noscript] attributes as hidden

Comment 7

[Mike McCabe]

Fix checked in.

Comment 8

[timeless]

Attachment: change nsISupportsVoid

Comment 9

[timeless]

Attachment: fix caller

Comment 10

[Doug Turner (:dougt)]

Comment on attachment 76244 [details] [diff] [review]
change nsISupportsVoid

r=dougt

Comment 11

[timeless]

Comment on attachment 76251 [details] [diff] [review]
fix caller

moved to bug 133657