Strict mode poision pills.

Strict mode poision pills.
 - function.caller
 - function.arguments

BUG=
TEST=test/mjsunit/strict-mode.js

[Martin Maly, 2011-03-14 05:00:15]

This change implements strict more poison pill functions for Function.arguments
and Function.caller for strict mode functions.

Solution is to create alternative function object maps for strict mode functions
which have custom callback pair for caller/arguments which always throws.

The arguments object poison pills coming soon (now that I know how to custom
craft v8 objects and their maps)

Thank you!
Martin

[Lasse Reichstein, 2011-03-14 12:20:14]

LGTM

http://codereview.chromium.org/6691003/diff/1/src/arm/codegen-arm.cc
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/6691003/diff/1/src/arm/codegen-arm.cc#newcode3118
src/arm/codegen-arm.cc:3118: !function_info->strict_mode()) {  // Strict mode
functions use slow path.
We'll need a fast path for strict mode functions too, eventually.

http://codereview.chromium.org/6691003/diff/1/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/6691003/diff/1/src/objects.cc#newcode5549
src/objects.cc:5549: map() ==
context()->global_context()->function_map_strict()) ||
Try splitting it into two asserts:
 ASSERT(!shared()->strict_mode() || map == ....->function_map_strict());
 ASSERT(shared()->strict_mode() || map == ....->function_map());
It's easier to see which one fails, if it does.

(Why isn't logical implication an operator :)

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js
File test/mjsunit/strict-mode.js (right):

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js#new...
test/mjsunit/strict-mode.js:991: 
How about an inherited strictness:
  var athird = (function() { "use strict"; return function(){};})();
or do we consider it well tested that this generates a strict function?

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js#new...
test/mjsunit/strict-mode.js:996: assertThrows(pill, TypeError);
Should you check whether it has a .prototype property?

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js#new...
test/mjsunit/strict-mode.js:1004: assertEquals(false, descriptor.configurable);
I think we have an assertFalse function.

[Martin Maly, 2011-03-14 16:39:26]

Thanks, Lasse. Made all changes. The fast path work is captured as an issue
1249.
The test you recommended - to check for prototype of the poison pill - revealed
bug (no prototype on the poison pill function) which was trivial to fix - set
the map of the poison pill function to the map with prototype. Since it was only
1 line change I am going proceeding to landing.

Thank you!
Martin

http://codereview.chromium.org/6691003/diff/1/src/arm/codegen-arm.cc
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/6691003/diff/1/src/arm/codegen-arm.cc#newcode3118
src/arm/codegen-arm.cc:3118: !function_info->strict_mode()) {  // Strict mode
functions use slow path.
On 2011/03/14 12:20:15, Lasse Reichstein wrote:
> We'll need a fast path for strict mode functions too, eventually.

Yep. Opening an issue on this.
http://code.google.com/p/v8/issues/detail?id=1249

http://codereview.chromium.org/6691003/diff/1/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/6691003/diff/1/src/objects.cc#newcode5549
src/objects.cc:5549: map() ==
context()->global_context()->function_map_strict()) ||
On 2011/03/14 12:20:15, Lasse Reichstein wrote:
> Try splitting it into two asserts:
>  ASSERT(!shared()->strict_mode() || map == ....->function_map_strict());
>  ASSERT(shared()->strict_mode() || map == ....->function_map());
> It's easier to see which one fails, if it does.
> 
> (Why isn't logical implication an operator :)
>  

Done.

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js
File test/mjsunit/strict-mode.js (right):

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js#new...
test/mjsunit/strict-mode.js:991: 
On 2011/03/14 12:20:15, Lasse Reichstein wrote:
> How about an inherited strictness:
>   var athird = (function() { "use strict"; return function(){};})();
> or do we consider it well tested that this generates a strict function?

Done.

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js#new...
test/mjsunit/strict-mode.js:996: assertThrows(pill, TypeError);
On 2011/03/14 12:20:15, Lasse Reichstein wrote:
> Should you check whether it has a .prototype property?

Good suggestion. Turns out there was a bug here. After I create the JSFunction
for poison pill I need to set its map to the strict map with prototype. Made
that 1 line change.

http://codereview.chromium.org/6691003/diff/1/test/mjsunit/strict-mode.js#new...
test/mjsunit/strict-mode.js:1004: assertEquals(false, descriptor.configurable);
On 2011/03/14 12:20:15, Lasse Reichstein wrote:
> I think we have an assertFalse function.

Done.