153830 - Mozilla 1.0's handling of null is different from IE, Netscape 4 and Netscape 6.

Status: RESOLVED FIXED

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

Description

[Huazhang Shao]

Try the following JS code on IE, Netscape 4, Netscape 6 and Mozilla 1.0 {Build
ID: 2002053012}
	var a=null;
	alert(a);
	var b="a="+a;
	alert(b);

There are two alert boxes.

On IE, Netscape 4 and Netscape 6
1) The first alert box shows
	null
2) The second alert box shows
	a=null

On Mozilla 1.0
1) The first alert box shows NOTHING.
2) The second alert box shows
	a=null

The difference has already caused errors in my code.

Comment 1

[CoL]

Confirmed on build: 2002062008 (win2k)

Comment 2

[Phil Schwartau]

Reassigning to DOM Level 0, as this seems to be an issue with 
the alert() function itself. There is no such problem with the
print() function in the standalone JS shell:

js> var a = null;
js> print(a);
null

js> var b="a="+a;
js> print(b);
a=null

Comment 3

[Phil Schwartau]

Attachment: Reporter's HTML testcase

Comment 4

[Johnny Stenback (:jst)]

Attachment: Special handling of alert(null) to make it show "null" in the alert

Comment 5

[Huazhang Shao]

I found a similar null-related incompatibility problem in escape().

Try the following JS code on IE, Netscape 4, Netscape 6 and Mozilla 1.0
	alert(typeof(escape(null))+'\n'+escape(null).length);

On IE, Netscape 4 and Netscape 6, the alert box shows
	string
	4
(escape(null) is 'null'.)

But on Mozilla 1.0, the alert box shows
	string
	0
(escape(null) is ''.)

Comment 6

[Johnny Stenback (:jst)]

You'll find this same difference in most methods that take string arguments. See
bug 69468 for more details on this.

Comment 7

[Peter Van der Beken [:peterv]]

Comment on attachment 89037 [details] [diff] [review]
Special handling of alert(null) to make it show "null" in the alert

>-  // XXX: Concatenation of optional args?
>+  // This should be done with a const nsAString&, but compilers suck
>+  // and can't deal with that :-(
>+  const nsAString *str = DOMStringIsNull(aString) ? &null_str : &aString;

I think you could do

  const nsAString& str = DOMStringIsNull(aString) ? NS_STATIC_CAST(const
nsAString&, null_str) : aString;

>@@ -5651,7 +5653,6 @@
>       STRING_TO_JSVAL(::JS_InternString(cx, "preferenceinternal"));
>   }
> 
>-  NS_ENSURE_SUCCESS(rv, rv);
>   PRUint32 action;
>   if (argc == 1) {
>       action = nsIXPCSecurityManager::ACCESS_GET_PROPERTY;

This change is obsolete I think.

Comment 8

[Johnny Stenback (:jst)]

Mass-reassigning bugs to [e-mail]

Comment 9

[Peter Van der Beken [:peterv]]

You have a patch and I sr'ed it with comments. Maybe you should check in? :-)

Comment 10

[Johnny Stenback (:jst)]

Comment on attachment 89037 [details] [diff] [review]
Special handling of alert(null) to make it show "null" in the alert

Hmm, yeah, I think I should :-)

I've heard casting to references is potentially bad on some compilers (can
cause copy ctors to be called n' what not), so I'll stay away from that. I just
removed the comment where I was whining about it in stead :-)

Comment 11

[Johnny Stenback (:jst)]

Comment on attachment 89037 [details] [diff] [review]
Special handling of alert(null) to make it show "null" in the alert

Trivial, and can help developers figure out what's going on. Requesting 1.7
approval.

Comment 12

[Asa Dotzler [:asa]]

Comment on attachment 89037 [details] [diff] [review]
Special handling of alert(null) to make it show "null" in the alert

a=asa (on behalf of drivers) for checkin to 1.7

Comment 13

[Johnny Stenback (:jst)]

Fix checked in.