Fix Array.prototype.{reduce,reduceRight} to pass undefined as receiver

Fix Array.prototype.{reduce,reduceRight} to pass undefined as receiver for strict mode callbacks.

Propagate strict mode information from pre-parser to parser for lazily compiled functions.

R=lrn@chromium.org
BUG=v8:1436
TEST=mjsunit/regress/regress-1436.js

Committed: http://code.google.com/p/v8/source/detail?r=8227

[Lasse Reichstein, 2011-06-08 18:41:42]

http://codereview.chromium.org/7044054/diff/1/src/array.js
File src/array.js (right):

http://codereview.chromium.org/7044054/diff/1/src/array.js#newcode1007
src/array.js:1007: if (%_CallFunction(receiver, current, i, this, f)) {
Does this correctly pass the global receiver to the function if "receiver" is
the undefined value?

http://codereview.chromium.org/7044054/diff/1/test/mjsunit/regress/regress-14...
File test/mjsunit/regress/regress-1436.js (right):

http://codereview.chromium.org/7044054/diff/1/test/mjsunit/regress/regress-14...
test/mjsunit/regress/regress-1436.js:46: assertEquals(void 0, [2,
3].reduce(function(){ return this; }));
That's incorrect. The "this" value should be the global object for a non-strict
function.

http://codereview.chromium.org/7044054/diff/1/test/mjsunit/regress/regress-14...
test/mjsunit/regress/regress-1436.js:56: [2, 3].map(function() {});
Try testing with both strict and non-strict functions, and check that they get
undefined and global object respectively.

[Mads Ager (chromium), 2011-06-09 07:13:34]

D'oh! I seem to remember having done this before. Thanks Lasse. :-)

New patch uploaded that actually fixes the issue. Strict mode information is now
also propagated from preparser to parser for lazily compiled functions.

[Lasse Reichstein, 2011-06-09 08:57:02]

LGTM

http://codereview.chromium.org/7044054/diff/4001/test/mjsunit/regress/regress...
File test/mjsunit/regress/regress-1436.js (right):

http://codereview.chromium.org/7044054/diff/4001/test/mjsunit/regress/regress...
test/mjsunit/regress/regress-1436.js:40: [2, 3].reduce(strict);
Group by function called, like below (i.e., both reduce calls before the
reduceRight calls).

http://codereview.chromium.org/7044054/diff/4001/test/mjsunit/regress/regress...
test/mjsunit/regress/regress-1436.js:54: [2, 3].map(strict);
How about putting them in alphabetic order.
Matbe also call with explicit thisArg values, both undefined and null (the
strict mode function would need to be different for the null call).

[Mads Ager (chromium), 2011-06-09 09:05:03]

http://codereview.chromium.org/7044054/diff/4001/test/mjsunit/regress/regress...
File test/mjsunit/regress/regress-1436.js (right):

http://codereview.chromium.org/7044054/diff/4001/test/mjsunit/regress/regress...
test/mjsunit/regress/regress-1436.js:40: [2, 3].reduce(strict);
On 2011/06/09 08:57:02, Lasse Reichstein wrote:
> Group by function called, like below (i.e., both reduce calls before the
> reduceRight calls).

Done.

http://codereview.chromium.org/7044054/diff/4001/test/mjsunit/regress/regress...
test/mjsunit/regress/regress-1436.js:54: [2, 3].map(strict);
On 2011/06/09 08:57:02, Lasse Reichstein wrote:
> How about putting them in alphabetic order.
> Matbe also call with explicit thisArg values, both undefined and null (the
> strict mode function would need to be different for the null call).

Done.