Support globals in frog isolates.

Support globals in frog isolates.

Committed: https://code.google.com/p/dart/source/detail?r=1667

[Siggi Cherem (dart-lang), 2011-11-18 01:28:09]

This also addresses a few of the comments raised after I submitted the previous
CL on isolates.

http://codereview.chromium.org/8591033/diff/5001/corelib/src/implementation/h...
File corelib/src/implementation/hash_map_set.dart (right):

http://codereview.chromium.org/8591033/diff/5001/corelib/src/implementation/h...
corelib/src/implementation/hash_map_set.dart:1: // Copyright (c) 2011, the Dart
project authors.  Please see the AUTHORS file
I don't plan to submit this file here - there is a separate code review for
this, which is currently under review.

http://codereview.chromium.org/8591033/diff/5001/frog/lib/isolate.dart
File frog/lib/isolate.dart (right):

http://codereview.chromium.org/8591033/diff/5001/frog/lib/isolate.dart#newcod...
frog/lib/isolate.dart:482: var src = script && script.src;
this was unrelated to this CL, but something that will unblock a user.

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart
File frog/member.dart (right):

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart#newcode407
frog/member.dart:407: if (declaringType.library == world.dom) {
I don't like all these subtle cases (here/below), let me know if you think of
something more general here. 

Basically we don't want to move natives into $globals, in the case of the DOM
this is not marked as a native field, maybe it should be?

[Jennifer Messerly, 2011-11-18 01:46:39]

lgtm

http://codereview.chromium.org/8591033/diff/5001/frog/gen.dart
File frog/gen.dart (right):

http://codereview.chromium.org/8591033/diff/5001/frog/gen.dart#newcode373
frog/gen.dart:373: writer.enterBlock('function \$static_init(){');
presumably library top-level variables need to go in here too?

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart
File frog/member.dart (right):

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart#newcode407
frog/member.dart:407: if (declaringType.library == world.dom) {
On 2011/11/18 01:28:10, sigmund wrote:
> I don't like all these subtle cases (here/below), let me know if you think of
> something more general here. 
> 
> Basically we don't want to move natives into $globals, in the case of the DOM
> this is not marked as a native field, maybe it should be?

yes, if it can be solved by marking native--that would be ideal.

I think the issue is that there's no way to put "native" on top level variables.
:\

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart#newcode765
frog/member.dart:765: declaringType.markUsed();
hmmm, we probably don't want to generate an entire type because one static
method is implicitly closure bound. However, I noticed that "invoke" has the
same issue. Can you add a TODO here and in MethodMember.invoke?

[Jennifer Messerly, 2011-11-18 01:47:03]

http://codereview.chromium.org/8591033/diff/5001/frog/gen.dart
File frog/gen.dart (right):

http://codereview.chromium.org/8591033/diff/5001/frog/gen.dart#newcode373
frog/gen.dart:373: writer.enterBlock('function \$static_init(){');
On 2011/11/18 01:46:39, John Messerly wrote:
> presumably library top-level variables need to go in here too?

Ignore this comment. I forgot to delete it :)

[jimhug, 2011-11-18 17:49:17]

LGTM!

On to web workers?

http://codereview.chromium.org/8591033/diff/5001/corelib/src/implementation/h...
File corelib/src/implementation/hash_map_set.dart (right):

http://codereview.chromium.org/8591033/diff/5001/corelib/src/implementation/h...
corelib/src/implementation/hash_map_set.dart:40: static final
_DeletedKeySentinel _DELETED_KEY = const _DeletedKeySentinel();
Nice!

http://codereview.chromium.org/8591033/diff/5001/frog/gen.dart
File frog/gen.dart (right):

http://codereview.chromium.org/8591033/diff/5001/frog/gen.dart#newcode25
frog/gen.dart:25: bool hasStatics = false;
FYI - Nice comment - this could also be very relevant if we decided to start
using a null sentinel object and thus may need to initialize some of these
fields to the sentinel.

http://codereview.chromium.org/8591033/diff/5001/tests/isolate/isolate.status
File tests/isolate/isolate.status (left):

http://codereview.chromium.org/8591033/diff/5001/tests/isolate/isolate.status...
tests/isolate/isolate.status:26: StaticStateTest: Fail
Yay!

[Siggi Cherem (dart-lang), 2011-11-18 18:33:17]

thanks!

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart
File frog/member.dart (right):

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart#newcode407
frog/member.dart:407: if (declaringType.library == world.dom) {
On 2011/11/18 01:46:39, John Messerly wrote:
> On 2011/11/18 01:28:10, sigmund wrote:
> > I don't like all these subtle cases (here/below), let me know if you think
of
> > something more general here. 
> > 
> > Basically we don't want to move natives into $globals, in the case of the
DOM
> > this is not marked as a native field, maybe it should be?
> 
> yes, if it can be solved by marking native--that would be ideal.
> 
> I think the issue is that there's no way to put "native" on top level
variables.
> :\

yeah :(

http://codereview.chromium.org/8591033/diff/5001/frog/member.dart#newcode765
frog/member.dart:765: declaringType.markUsed();
On 2011/11/18 01:46:39, John Messerly wrote:
> hmmm, we probably don't want to generate an entire type because one static
> method is implicitly closure bound. However, I noticed that "invoke" has the
> same issue. Can you add a TODO here and in MethodMember.invoke?

Sure. Done.