Constructor tear-off closures

Implement constructor closures

One detail that isn't implemented yet is canonicalization of constructor tear-offs. At the moment, two tear-offs of the same constructor result in closures that are not equal.

BUG=
R=regis@google.com

Committed: https://github.com/dart-lang/sdk/commit/4ebab25d8f250f389a991148ba847c2b5d3c7c04

[hausner, 2015-08-03 20:14:50]

hausner@google.com changed reviewers:
+ srdjan@google.com

[hausner, 2015-08-03 20:32:11]

hausner@google.com changed reviewers:
+ regis@google.com

[hausner, 2015-08-03 20:32:49]

Adding Regis who knows a thing or two about type parameters.

[regis, 2015-08-03 21:51:36]

LGTM

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc
File runtime/vm/parser.cc (right):

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:837: // of functions, special cases for getters and posibly
other kinds
possibly

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:1340: UNREACHABLE();
Is this leftover debug code?

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:12880: if (!sig_type.IsFinalized()) {
Better make sure we are not top level parsing, or finalization would be
premature:
ASSERT(!is_top_level_);

[hausner, 2015-08-03 22:51:01]

Committed patchset #3 (id:40001) manually as
4ebab25d8f250f389a991148ba847c2b5d3c7c04 (presubmit successful).

[hausner, 2015-08-03 22:52:16]

Thank you!

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc
File runtime/vm/parser.cc (right):

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:837: // of functions, special cases for getters and posibly
other kinds
On 2015/08/03 21:51:35, regis wrote:
> possibly

Done.

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:1340: UNREACHABLE();
On 2015/08/03 21:51:35, regis wrote:
> Is this leftover debug code?

It's a case that should never happen, because the formal parameters have been
parsed (and syntactically checked) before so parsing them again must succeed.
Changed to use an ASSERT and a USE() annotation to avoid build errors in release
mode.

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:12880: if (!sig_type.IsFinalized()) {
On 2015/08/03 21:51:35, regis wrote:
> Better make sure we are not top level parsing, or finalization would be
> premature:
> ASSERT(!is_top_level_);

Done.

[srdjan, 2015-08-03 23:27:30]

lgtm

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc
File runtime/vm/parser.cc (right):

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:847: const Script& script = Script::Handle(func.script());
Handle(isolate(), ...)

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:857:
Thread::Current()->isolate()->object_store()->clear_sticky_error();
Parser has isolate_ field:
isolate()->object_store()->clear....

[hausner, 2015-08-03 23:52:21]

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc
File runtime/vm/parser.cc (right):

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:847: const Script& script = Script::Handle(func.script());
On 2015/08/03 23:27:30, srdjan wrote:
> Handle(isolate(), ...)

This is in a static method, so isolate() is not available.

https://codereview.chromium.org/1255063005/diff/40001/runtime/vm/parser.cc#ne...
runtime/vm/parser.cc:857:
Thread::Current()->isolate()->object_store()->clear_sticky_error();
On 2015/08/03 23:27:30, srdjan wrote:
> Parser has isolate_ field:
> isolate()->object_store()->clear....

same here.