Set type argument vector at run time in instantiated closure objects.

Set type argument vector at run time in instantiated closure objects.
Set type argument vector at compile time in type of closure variables.
TODO: Set type argument vector at compile time in type of closure parameters.
Committed: https://code.google.com/p/dart/source/detail?r=416

[siva, 2011-10-13 20:52:21]

LGTM

http://codereview.chromium.org/8271008/diff/1/runtime/vm/code_generator_ia32.cc
File runtime/vm/code_generator_ia32.cc (right):

http://codereview.chromium.org/8271008/diff/1/runtime/vm/code_generator_ia32....
runtime/vm/code_generator_ia32.cc:914: __ popl(ECX);  // Pop receiver.
The code for VisitClosureNode and VisitImplicitInstanceClosureNode seem
identical in the middle, I am wondering if it be worth making a static method of
the common code.

http://codereview.chromium.org/8271008/diff/1/runtime/vm/code_generator_ia32....
runtime/vm/code_generator_ia32.cc:2250: __ popl(EAX);
// Pop instantiator.

http://codereview.chromium.org/8271008/diff/1/runtime/vm/stub_code_ia32.cc
File runtime/vm/stub_code_ia32.cc (right):

http://codereview.chromium.org/8271008/diff/1/runtime/vm/stub_code_ia32.cc#ne...
runtime/vm/stub_code_ia32.cc:1247: //   ESP : points to return address.
if signature class is not parameterized  ESP + 4 will be the receiver right,
maybe we should document that?

http://codereview.chromium.org/8271008/diff/1/runtime/vm/stub_code_ia32.cc#ne...
runtime/vm/stub_code_ia32.cc:1368: __ PushObject(TypeArguments::ZoneHandle()); 
// Push null type arguments.
why not use pushl(raw_null) here?

[regis, 2011-10-13 21:33:46]

Thanks!

http://codereview.chromium.org/8271008/diff/1/runtime/vm/code_generator_ia32.cc
File runtime/vm/code_generator_ia32.cc (right):

http://codereview.chromium.org/8271008/diff/1/runtime/vm/code_generator_ia32....
runtime/vm/code_generator_ia32.cc:914: __ popl(ECX);  // Pop receiver.
On 2011/10/13 20:52:21, asiva wrote:
> The code for VisitClosureNode and VisitImplicitInstanceClosureNode seem
> identical in the middle, I am wondering if it be worth making a static method
of
> the common code.

I tried to factorize the visiting code of all 3 closure flavors, but it gets
harder to read.

Even for the 2 instance flavors, many parameters need to be passed to the common
function, since the 2 ast node types are unrelated.

At this point, it would make sense to consolidate all 3 ast nodes into a single
one, For now. I will leave it as is.

Added TODO.

http://codereview.chromium.org/8271008/diff/1/runtime/vm/code_generator_ia32....
runtime/vm/code_generator_ia32.cc:2250: __ popl(EAX);
On 2011/10/13 20:52:21, asiva wrote:
> // Pop instantiator.

Done.

http://codereview.chromium.org/8271008/diff/1/runtime/vm/stub_code_ia32.cc
File runtime/vm/stub_code_ia32.cc (right):

http://codereview.chromium.org/8271008/diff/1/runtime/vm/stub_code_ia32.cc#ne...
runtime/vm/stub_code_ia32.cc:1247: //   ESP : points to return address.
On 2011/10/13 20:52:21, asiva wrote:
> if signature class is not parameterized  ESP + 4 will be the receiver right,
> maybe we should document that?

Done.

http://codereview.chromium.org/8271008/diff/1/runtime/vm/stub_code_ia32.cc#ne...
runtime/vm/stub_code_ia32.cc:1368: __ PushObject(TypeArguments::ZoneHandle()); 
// Push null type arguments.
On 2011/10/13 20:52:21, asiva wrote:
> why not use pushl(raw_null) here?

Done here and at some other locations.