22807 - [4.xP]Variables forgotten between SCRIPT elements

Status: VERIFIED DUPLICATE of bug 21618

(Core :: JavaScript Engine, defect, P3)

Description

[Joseph Elwell]

BrowserSpy does a good job of pointing out subtle differences between 4.x and
5.0

This example shows that mozilla seeming does not support Javascript versions 1.1
or higher.
The offending code always looks like this:
<SCRIPT LANGUAGE="JavaScript1.1">
<!--
ver11 = 1;
// -->
</SCRIPT>
<SCRIPT LANGUAGE="JavaScript" TYPE="text/javascript">
<!--
var ver11;
(ver11) ? document.writeln("Supported") : document.writeln("Not supported");
// -->
</SCRIPT>

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

The problem here isn't what you think it is.  If you remove the "1.1", you still
see "Not Supported".

The problem has to do with variables disappearing.  Normally, it seems that if
you have (within one SCRIPT element):

number = 1;
// number is 1
var number;
// number is still 1

If, you define a variable "number" in one SCRIPT element, and then have a "var
number" in another SCRIPT element, then the variable defined in the first
element doesn't show up in the second (it's undefined), even before the "var
number"!  (However, without that "var number", things work fine.)

I'm not sure from the symptoms whether this is Core JavaScript or DOM (and I do
understand the difference at some abstract level).  Assigning to JavaScript
Engine component and reassigning.

I'll attach testcases showing the problem later this evening.

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: testcase showing bug

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: similar testcase, not showing bug

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

For the record, I'm using Win95 mozilla 1999-12-23-09-M13.

Comment 5

[Mike McCabe]

Is this a new bug?  If so, then there may be something to fix.

I'm guessing not, though; the behavior you describe is how var statements work.

Each script tag is logically a separate script, though subsequent tags operate
on a a global object that has already had names and properties added by previous
script tags.  Within a given script or function body, the way var statements
work is that the 'var' part of the statement is implicitly promoted to the top
of the *script* or function (not top of page; only within the current script
tag.) and the associated variable is given the undefined value at that time.
Looks like that's overwriting existing values for that name from previous script
tags.  Then, if there's an initializer ('= 4') associated with the var
statement, that executes at the point in the script where the var statement
occurs.

It's all counter-intuitive, but thanks to the ECMA standards process, it's also
all defined behavior.

Please go ahead and close this bug, *unless* this behavior is new to Mozilla, in
which case there's probably something else going on.

Comment 6

[Mike McCabe]

Another thought -

I'm not sure if the ECMA-262 spec for JavaScript covers the case where multiple
scripts (i.e. script tags) work on the same object.  So again, it'd be
interesting to know if it's a change from earlier behavior.  If it is, it could
have something to do with how/when we initialize with the undefined value.  If
this is existing behavior, then it's probably one of those subtle aspects of
behavior that's too dangerous to change.

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

This is a new bug as compared to 4.x.  The first testcase shows "Number is 1"
three times in Linux NN 4.7, but "number is 1 / number is undefined / number is
undefined" on Mozilla 2000-01-04-09-M13 linux.

Changing to All/All from PC/Windows NT.

Comment 8

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Also, although I haven't looked at the ECMAScript spec for quite a while (and
when I did I wasn't really looking at the parts relevant to this), I find it
hard to believe that the contents of the second script element can influence
what variables are defined at the beginning of execution of that element.

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Well, after reading your explanation in more detail, it's not so hard to
believe, but there's still a 4.xP bug here.

Comment 10

[Mike McCabe]

Promotion of var statements to the top of the script or function body happens
before any execution occurs.  Just how this happens could be the issue - looks
like this bug could be 'variables are assigned the undefined value rather than
inheriting their value from previous script invocations.'

Could have something to do with how we're doing scopes.  Reassigning to Roger
Lawrence.

Comment 11

[ekrock's old account (dead)]

Bulk moving [testcase] code to new testcase keyword. Sorry for the spam!

Comment 12

[rogerl (gone)]

Dup of #21618 - Brendan's multi-scripts in a page thing, now fixed.

*** This bug has been marked as a duplicate of 21618 ***

Comment 13

[Neil Marshall]

Verified dupe of 21618