Implement basic code generation for arrow functions

Implement basic code generation for arrow functions

Implements code generation for arrow functions by desugaring them into
a FunctionLiteral. For the moment, a normal FUNCTION_SCOPE is used, so
"this" and "arguments" behave as in normal functions. Implementing the
correct scoping rules is to be done later on.

BUG=v8:2700
LOG=
R=rossberg@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=22495

[rossberg, 2014-07-14 13:55:47]

LGTM, only nits

https://codereview.chromium.org/382893003/diff/1/src/objects.h
File src/objects.h (right):

https://codereview.chromium.org/382893003/diff/1/src/objects.h#newcode7279
src/objects.h:7279: // Indicated that this function is an arrow function.
s/Indiciated/Indicates/

https://codereview.chromium.org/382893003/diff/1/src/v8natives.js
File src/v8natives.js (right):

https://codereview.chromium.org/382893003/diff/1/src/v8natives.js#newcode1737
src/v8natives.js:1737: return source;
Nit: please use same brace style as surrounding code

https://codereview.chromium.org/382893003/diff/1/test/cctest/test-parsing.cc
File test/cctest/test-parsing.cc (right):

https://codereview.chromium.org/382893003/diff/1/test/cctest/test-parsing.cc#...
test/cctest/test-parsing.cc:256: NULL
Nit: Don't use NULL, use the ARRAY_SIZE macro instead.

https://codereview.chromium.org/382893003/diff/1/test/mjsunit/harmony/arrow-f...
File test/mjsunit/harmony/arrow-functions.js (right):

https://codereview.chromium.org/382893003/diff/1/test/mjsunit/harmony/arrow-f...
test/mjsunit/harmony/arrow-functions.js:9: assertEquals(typeof (() => {}),
"function");
Nit: expected value should go left (similarly below a few times).

https://codereview.chromium.org/382893003/diff/1/test/mjsunit/harmony/arrow-f...
test/mjsunit/harmony/arrow-functions.js:46: [1, 2, 3, 4, 5, 6, 7, 8, 9,
10].forEach(v => { if (v % 5 === 0) fives.push(v); });
Nit: line length

[aperez, 2014-07-14 18:20:40]

On 2014/07/14 13:55:47, rossberg wrote:
> LGTM, only nits

The new patch set should have all the nits fixed. As I am not a committer, could
you please land this for me?

[rossberg, 2014-07-16 09:37:41]

Hm, I'm getting this error:

../src/ast.h:2435:56: error: ‘body’ may be used uninitialised in this function
[-Werror=uninitialized]
../src/preparser.h:2484:40: note: ‘body’ was declared here

[aperez, 2014-07-16 11:18:52]

On 2014/07/16 09:37:41, rossberg wrote:
> Hm, I'm getting this error:
> 
> ../src/ast.h:2435:56: error: ‘body’ may be used uninitialised in this function
> [-Werror=uninitialized]
> ../src/preparser.h:2484:40: note: ‘body’ was declared here

Ah, let me do a build with GCC and werror=yes — with my version of Clang here
I need to pass werror=no and sometimes I miss things :-/

[aperez, 2014-07-16 11:42:14]

On 2014/07/16 11:18:52, aperez wrote:

> Ah, let me do a build with GCC and werror=yes — with my version of Clang here
> I need to pass werror=no and sometimes I miss things :-/

With this I have also catched a signed vs. unsigned comparison in
“test-parsing.cc”. The last version of the patchset builds now fine
with “werror=yes” and it should be good to land.

[rossberg, 2014-07-16 11:59:42]

On 2014/07/16 11:42:14, aperez wrote:
> On 2014/07/16 11:18:52, aperez wrote:
> 
> > Ah, let me do a build with GCC and werror=yes — with my version of Clang
here
> > I need to pass werror=no and sometimes I miss things :-/
> 
> With this I have also catched a signed vs. unsigned comparison in
> “test-parsing.cc”. The last version of the patchset builds now fine
> with “werror=yes” and it should be good to land.

Hm, does the test suite run through for you? I see plenty of tests failing right
now.

[aperez, 2014-07-19 20:33:27]

On 2014/07/16 11:59:42, rossberg wrote:
> On 2014/07/16 11:42:14, aperez wrote:
> > On 2014/07/16 11:18:52, aperez wrote:
> > 
> > > Ah, let me do a build with GCC and werror=yes — with my version of Clang
> here
> > > I need to pass werror=no and sometimes I miss things :-/
> > 
> > With this I have also catched a signed vs. unsigned comparison in
> > “test-parsing.cc”. The last version of the patchset builds now fine
> > with “werror=yes” and it should be good to land.
> 
> Hm, does the test suite run through for you? I see plenty of tests failing
right
> now.

Fixed, it ws two things: I missed adding FunctionisArrow to the list of runtime
functions in “runtime.h”. Also, I have changed a couple of cases in which the
code
was like:

   return Blah(CHECK_OK);

to:

   result = Blah(CHECK_OK);
   return result;

This makes sure that the CHECK_OK macro is expanded in a way that makes sense.

[aperez, 2014-07-19 20:33:35]

On 2014/07/19 20:33:27, aperez wrote:
> On 2014/07/16 11:59:42, rossberg wrote:
> > On 2014/07/16 11:42:14, aperez wrote:
> > > On 2014/07/16 11:18:52, aperez wrote:
> > > 
> > > > Ah, let me do a build with GCC and werror=yes — with my version of Clang
> > here
> > > > I need to pass werror=no and sometimes I miss things :-/
> > > 
> > > With this I have also catched a signed vs. unsigned comparison in
> > > “test-parsing.cc”. The last version of the patchset builds now fine
> > > with “werror=yes” and it should be good to land.
> > 
> > Hm, does the test suite run through for you? I see plenty of tests failing
> right
> > now.
> 
> Fixed, it ws two things: I missed adding FunctionisArrow to the list of
runtime
> functions in “runtime.h”. Also, I have changed a couple of cases in which the
> code
> was like:
> 
>    return Blah(CHECK_OK);
> 
> to:
> 
>    result = Blah(CHECK_OK);
>    return result;
> 
> This makes sure that the CHECK_OK macro is expanded in a way that makes sense.

[aperez, 2014-07-19 20:34:22]

On 2014/07/19 20:33:35, aperez wrote:
> On 2014/07/19 20:33:27, aperez wrote:
> > On 2014/07/16 11:59:42, rossberg wrote:
> > > On 2014/07/16 11:42:14, aperez wrote:
> > > > On 2014/07/16 11:18:52, aperez wrote:
> > > > 
> > > > > Ah, let me do a build with GCC and werror=yes — with my version of
Clang
> > > here
> > > > > I need to pass werror=no and sometimes I miss things :-/
> > > > 
> > > > With this I have also catched a signed vs. unsigned comparison in
> > > > “test-parsing.cc”. The last version of the patchset builds now fine
> > > > with “werror=yes” and it should be good to land.
> > > 
> > > Hm, does the test suite run through for you? I see plenty of tests failing
> > right
> > > now.
> > 
> > Fixed, it ws two things: I missed adding FunctionisArrow to the list of
> runtime
> > functions in “runtime.h”. Also, I have changed a couple of cases in which
the
> > code
> > was like:
> > 
> >    return Blah(CHECK_OK);
> > 
> > to:
> > 
> >    result = Blah(CHECK_OK);
> >    return result;
> > 
> > This makes sure that the CHECK_OK macro is expanded in a way that makes
sense.

BTW, all test cases pass, including now “test262” and the “mozilla” suites.

[rossberg, 2014-07-21 09:58:15]

Committed patchset #4 manually as r22495 (presubmit successful).