sparse array trick fails on Opera 10

[Utz-Uwe Haus]

Opera 10 (haven't checked older versions) un-sparsifies sparse arrays on copying to have `undefined' entries in the previously unused slots. This breaks the code generated by dojo._listener.getDispatcher(). For me changing the

if(!dojo.isRhino)

to

if(!dojo.isRhino && !dojo.isOpera)

fixes the problem. Apparently the surrounding includeStart-magic should be updated too.

Unfortunately I have no small test case to attach; actually I have no clear idea how to trigger the situation that the handler array becomes sparse. However, try [].concat(ls) on a sparse `ls' exhibits the problem.

This is similar to the Rhino special casing done in #7523.

[bill]

Same as #10010?

[Utz-Uwe Haus]

Seems very much so.

[bill]

The issue is apparently with setting up multiple handlers on a *plain function*, as opposed to handlers for DOM events like onclick. And the sparse array issue apparently comes from disconnecting a handler.

The code as-is is working fine for me on Opera10/mac and Opera10/windows. I'll attach my test case. I'm not seeing gaps in the array.

My opera is:

Version
10.00

Build
6652

[bill]

works fine for me, put this file parallel to dojo/ and dijit/ directories

[Sami]

Could reproduce the issue while trying to use both dojo and prototype.js on the same page. When prototype.js is included into bill's testpage the array becomes sparse and opera 10 throws an error (see attachment).

[Sami]

reproducing sparse array byusing both dojo and prototype.js on same page

[Utz-Uwe Haus]

I can confirm that my initial report was from a scenario using prototype.js and dojo on the same page too.

[bill]

OK, I see it too, with ​prototype 1.6.1.

Maybe we should just make the rhino path the default, getting rid of the current default path.

[cweiske]

I see this bug in Opera 11.00 when using Tiny Tiny RSS (​http://tt-rss.org/redmine/issues/291), so it is still relevant at least for me. Any fix in sight?

[bill]

Note: the rhino code was removed in [22570] and merged to trunk in [22989].

I tried sparse arrays on opera 10 with prototype loaded (from the page ​http://www.prototypejs.org/), they seem to work fine:

>>> var ary = ['a', 'b', 'c'];delete ary[1];var copy = [].concat(ary); 
true
>>> copy[1] 
undefined

On opera11 it also worked as expected for me, from the address bar:

javascript:var ary = ['a', 'b', 'c'];delete ary[1];var copy = [].concat(ary); alert(copy[1]);

So requires some more investigation.

[hallvord@…]

This is caused by a Prototype workaround against an earlier Opera bug/incompatibility.

As a result of this workaround, array.concat() doesn't work as expected on sparse arrays:

if (Prototype.Browser.Opera){
  Array.prototype.concat = function() {
    var array = [];
    for (var i = 0, length = this.length; i < length; i++) array.push(this[i]);
    for (var i = 0, length = arguments.length; i < length; i++) {
      if (Object.isArray(arguments[i])) {
        for (var j = 0, arrayLength = arguments[i].length; j < arrayLength; j++)
          array.push(arguments[i][j]);
      } else {
        array.push(arguments[i]);
      }
    }
    return array;
  };
}

See ​https://prototype.lighthouseapp.com/projects/8886/tickets/972-issue-regarding-arrayconcat

[hallvord@…]

(BTW, I believe this bug is fixed in new Prototype.js versions, though I don't know exactly when the fix went out)

[Kris Zyp]

So this is a Prototype bug? I assume we can mark this bug as invalid then.

[hallvord@…]

yes to both questions :)