Execuse me. It doesn't seem to be makeClass, rather it's that 
_deserializeHeaders just doesn't exist as a member of SipSIPURI (see 
SipSyntaxFactory.js line 695). Why is it called?

Filip

Well ... it was a version conflict.

[...]

The problem here is that the metafunction SipSyntaxObject.parsedHash() 
discriminates between objects and regexps based on what 'typeof' 
returns. According to the JS specs 'typeof a' should return 'function' 
if a is a regexp and 'object' if a is an object.

Unfortunately this behaviour was recently silently changed. Mozilla now 
reports 'object' for regexps and objects. See bug#289933 
http://bugzilla.mozilla.org/show_bug.cgi?id=289933.

I've reopened the bug and asked for the change to be backed out, but 
I'll have a workaround ready in a bit.

- Alex

Hi Alex,

Alex Fritze schrieb:[...]

in my point of view, the typeof operator in ECMAScript isn't very 
reliable anyways, as strings or regexps are just objects with a special 
behaviour. For example:
  typeof "hello" == "string"
  typeof new String("hello") == "string"
but
  typeof function() {} == typeof new Function("", "") == "function".

For your case, wouldn't it be a better solution to check for (a 
instanceof RegExp)?

Filip

Filip Dalüge wrote:
[...]

The ECMA specs are _relatively_ clear on this. If an object implements 
[[Call]], typeof should return 'function'.
[...]
that our module system introduces.
E.g.:
----------------------------------------------
File A.js :
MOZ_EXPORT=[aregexpA, RegExpA];

var aregexpA = /foo/;
var RegExpA = RegExp;
----------------------------------------------
----------------------------------------------
File B.js:
importModule("A.js");
instanceof aregexpA == RegExp;  // 1.
instanceof aregexpA == RegExpA; // 2.
RegExp == RegExp; // 3.
----------------------------------------------

Each module gets its own set of the standard classes, so
expressions 1 & 3 will evaluate to 'false',
expression 2 will evaluate to 'true'.
This behaviour (which has its good and bad points) is likely to change 
at some point in the future though.

In any case, I have got a workaround which will be checked in within the 
hour.
Basically, the object/regexp discrimination code was used to make some 
parameters in the parsedAttrib and parsedHash meta-functions optional, 
but I have now inserted the optional arguments everywhere and backed out 
the offending code.

- Alex

[...]

This problem is now fixed in CVS.

- Alex

Alex Fritze schrieb:
[...]

OK, but there are some other rules that may lead to confusing cases in 
connection with typeof:
- null is of type "object" but throws an Exception when a script 
accesses its properties
- the type of a string depends on wether it has been created by the new 
operator (type "object") or it is represented as a string value (type 
"string") (that's what I meant to express in my last post but I mistyped)
- non-native objects may claim to be an object although they don't 
implement the properties of the Object prototype object

So I for myself prefer instanceof where possible.
[...]

As the imported module's global object is accessible, someone can also 
access the non-native standard classes:

var mod = importModule("A.js");
if(mod.aregexpA == instanceof mod.RegExp) /*...*/;

That's fine :-)
[...]

Great, thanks!

Filip

Marc Petit-Huguenin wrote:
[...]
bugzilla bug on this.
Have you tried compiling xulrunner or firefox?

- Alex