90325 - In the DOM , variable name 'prefix' causes problems in JavaScript with blocks

Status: RESOLVED INVALID

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

Description

[deneen]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9.2+)
Gecko/20010711
BuildID:    2001071104

variables defined outside of with(...){} are not accessable inside of the
with(...){}.

Reproducible: Always
Steps to Reproduce:
test case shows it

Actual Results:  variable does not exist

Expected Results:  an the alert messages should be the same

Comment 1

[deneen]

Attachment: test case for bug

Comment 2

[Axel Hecht]

Tested on solaris, confirming.
I tested this in xpcshell:
a=5;
with (Math) {a+random()};
which works, so it's not a general problem of with(), but maybe something 
related to xpconnect?

Axel

Comment 3

[David Bradley]

It appears to be DOM specific. If I take that same page place Math instead of
the DOM object. It also works.

I'll take a look and see what I can find. jst, brendan I copied you in case you
might know something off the top of your heads.

Comment 4

[David Bradley]

Also did a test with a non-DOM wrapped native and it worked as expected.

Comment 5

[Phil Schwartau]

Added testcase to JS test suite:

            mozilla/js/tests/js1_5/Scope/scope-004.js

This testcase passes in the standalone JS shell on all platforms.
Seeing no problems with this in the JS shell - 

Comment 6

[Phil Schwartau]

Attachment: HTML version of JS shell testcase

Comment 7

[Phil Schwartau]

The problem turns out to be the use of the variable name 'prefix' !!!

I will attach an HTML testcase showing this. It contains two links.
One runs a reduced version of [e-mail]'s testcase above, 
which used the variable name 'prefix'. This link shows the problem.

The other link runs the same testcase using the variable name 'prefi'
instead of 'prefix'. There is NO PROBLEM in this case -

Comment 8

[Phil Schwartau]

Attachment: HTML testcase comparing variable names 'prefix' vs. 'prefi'

Comment 9

[Phil Schwartau]

In ECMA-262, I do not see the word 'prefix' marked as a reserved word 
or as a future reserved word. Additionally, I see no problems in the 
standalone JS shell using it as an identifier in the testcases above.

Therefore reassigning this to DOM Level 0; it is not a JS Engine bug.
Is 'prefix' a reserved word in the DOM?

Comment 10

[Phil Schwartau]

Resummarizing - 

Comment 11

[Johnny Stenback (:jst)]

"prefix" is a property of all DOM nodes, it's the namespace prefix of a node
(i.e. the element "<foo:bar>" would have prefix="foo"), you'll haveto use a
different variable name. INVALID.

Comment 12

[Brendan Eich [:brendan]]

This is yet another reason why with statements are to be avoided in JS.  I'm
sorry I added the silly thing to the language.

/be

Comment 13

[deneen]

sorry about this....

prefix is a very common word, and had it not worked outside the with(), I would 
have figured it was a reserved word.

Is there no default value for prefix?  (so an alert(element.prefix) would say 
something other than undefined)

Comment 14

[Axel Hecht]

well, the prefix is the prefix, so as long as you're in the default namespace
(which is what mozilla assumes for HTML), your prefix is "".
It's just the way it is.

And of course a lol to brendan ;-)

Axel