String.prototype.replace works incorrectly with RegExp assertion ?=

[YLeight]

In case when we have regular expression with assertion ?=, the wrappedReplaceValue, which decorate the original replaceValue function, can't define correct arguments, because the coincidence which was passed to replaceValue function don't match the RegExp pattern and the final arguments have just only two values - the offset of the matched sub-string within the whole string being examined, and the whole string being examined.
There is a certain problem, when whole RegExp don't match with found coincidence, but in fact, there is a coincidence StringPrototype.replace polyfill

[ljharb]

Can you provide specific code (like a specific regex that has a lookahead)?

[YLeight]

/(\n)?(^[ \t]*)([*+-]|\d+[.])[ \t]+((\[(x|X| )?])?[ \t]*[^\r]+?(\n{1,2}))(?=\n*(~0|\2([*+-]|\d+[.])[ \t]+))/gm

[YLeight]

StringPrototype.replace polyfill
If the searchValue equal /(\n)?(^[ \t]*)([*+-]|\d+[.])[ \t]+((\[(x|X| )?])?[ \t]*[^\r]+?(\n{1,2}))(?=\n*(~0|\2([*+-]|\d+[.])[ \t]+))/gm and match equal '* some list item;' we get null in 1892 code line but not the groups.

[ljharb]

var str = '* some list item;';
var re = /(\n)?(^[ \t]*)([*+-]|\d+[.])[ \t]+((\[(x|X| )?])?[ \t]*[^\r]+?(\n{1,2}))(?=\n*(~0|\2([*+-]|\d+[.])[ \t]+))/gm;
[str.match(re), re.exec(str), str.replace(re, function () { console.log(arguments); })]

in the native browser console implies there should be no match. Can you provide a specific test that passes on a compliant browser but fails on a browser where the es5-shim has polyfilled String#replace?

[YLeight]

I mean next case:

var reg = /(\n)?(^[ \t]*)([*+-]|\d+[.])[ \t]+((\[(x|X| )?])?[ \t]*[^\r]+?(\n{1,2}))(?=\n*(~0|\2([*+-]|\d+[.])[ \t]+))/gm;
var str = "+ first item in list;\n\n+ second item in list;\n~0";

str.replace(reg, function () {
    console.log(arguments.length);
});

For example, firefox version 24:

• with es5-shim: arguments.length will be 2;
• without es5-shim: arguments.length will be 12;

[ljharb]

Thanks, are there specific browsers you're getting 2 in?

[YLeight]

I am shure, firefox version 24 is one of them.
And I assume, all firefox versions untill 33 act the same.
Gecko-specific notes (in bottom of article)
RegExp.$N mozilla bug - checking in