40760 - JS1.5 getters and setters may replace native properties

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect, P3)

Description

[Brendan Eich [:brendan]]

Waldemar's write-up said that it should be "verboten", e.g., to replace the 
lastIndex property of a regular expression with a getter or setter.  We do 
protect permanent properties, but not readonly and not "native" -- there is no 
internal attribute for such properties.

Suggestions on how to implement this are welcome.

/be

Comment 1

[Brendan Eich [:brendan]]

JSPROP_FINAL seems like the right thing.  Mega-patch coming up.

/be

Comment 2

[Brendan Eich [:brendan]]

Setting URL to the deja.com remembrance of waldemar's posting.

Thinking about this more, does it behoove us to protect properties that lack a 
DontDelete attribute from being overridden by a getter and/or setter?  Suppose 
we make re.lastIndex final -- one could still delete re.lastIndex and redefine 
it with a getter and setter that do non-standard magic.

/be

Comment 3

[Waldemar Horwat]

re.lastIndex is DontDelete according to ECMA v3.

Comment 4

[Brendan Eich [:brendan]]

Ok (lastIndex not being DontDelete is an ECMA ed. 3 bug that slipped past the 
testsuite, I guess; I'll fix it).

But are there any properties that lack DontDelete, but that should be protected 
from getter/setter replacement?  Or are there properties that are not DontDelete 
in ECMA yet, but that (given extensions __defineGetter__ and __defineSetter__) 
should now be DontDelete?

/be

Comment 5

[Brendan Eich [:brendan]]

Attachment: proposed partial fix

Comment 6

[Brendan Eich [:brendan]]

This patch fixes and externs js_CheckRedeclartion so it can be used from 
obj_defineGetter and obj_defineSetter (the native methods for __defineGetter__ 
and __defineSetter__).

It also makes RegExp.prototype.lastIndex PERMANENT, but what about all those 
other properties (ignoreCase, source, etc.)?  I don't have ECMA ed. 3 handy 
right now.  Phil, can you update the testsuite and report bogusly-deletable 
properties that ECMA says should be DontDelete?

/be

Comment 7

[Brendan Eich [:brendan]]

I found an E262-3.pdf, and notice that the following are DontDelete, but are not 
JSPROP_PERMANENT in SpiderMonkey:

12.14:  catch variables
15.1.1.1: NaN in the global object
15.1.1.2: Infinity in the global object
15.1.1.3: undefined in the global object
15.3.5.2: f.prototype for a function f
15.5.5.1: s.length for a string s
15.10.7: all properties of regular expression objects

That's it.  I'll have an updated patch shortly.  Phil, anyone, double-check me 
if you please.  JS1.5 is supposed to comply with ECMA-262 Edition 3.

/be

Comment 8

[Brendan Eich [:brendan]]

Here comes the new patch.  It's bigger than I'd like, because I had to add a new 
bytecode for catch-variable setup.  Cc'ing shaver.

/be

Comment 9

[Brendan Eich [:brendan]]

Attachment: proposed total fix

Comment 10

[Mike Shaver (:shaver emeritus)]

While you're in jsopcode.c, can you fix the decompilation bug?

 		    len = js_GetSrcNoteOffset(sn, 0);
 		    pc += 4;	/* initprop, enterwith */
 		    if (len) {
-			js_printf(jp, " : ");
+			js_printf(jp, " if ");

Other than that, it looks good.

Comment 11

[Brendan Eich [:brendan]]

Attachment: revised proposed total fix

Comment 12

[Brendan Eich [:brendan]]

Fix checked in.

/be