Don't assume that a transform node is consistently deferred or not.

Don't assume that a transform node is consistently deferred or not.

R=rnystrom@google.com

Committed: https://code.google.com/p/dart/source/detail?r=35606

[Bob Nystrom, 2014-04-29 23:46:21]

https://codereview.chromium.org/255283003/diff/1/pkg/barback/test/package_gra...
File pkg/barback/test/package_graph/declaring_transformer_test.dart (right):

https://codereview.chromium.org/255283003/diff/1/pkg/barback/test/package_gra...
pkg/barback/test/package_graph/declaring_transformer_test.dart:224: test("still
runs apply", () {
This seems like a perverse edge case to add complexity to support. If a
transformer fails when declaring its outputs, why apply it?

[nweiz, 2014-04-30 00:36:57]

https://codereview.chromium.org/255283003/diff/1/pkg/barback/test/package_gra...
File pkg/barback/test/package_graph/declaring_transformer_test.dart (right):

https://codereview.chromium.org/255283003/diff/1/pkg/barback/test/package_gra...
pkg/barback/test/package_graph/declaring_transformer_test.dart:224: test("still
runs apply", () {
On 2014/04/29 23:46:22, Bob Nystrom wrote:
> This seems like a perverse edge case to add complexity to support. If a
> transformer fails when declaring its outputs, why apply it?

Why not apply it? As far as we know, it'll still function perfectly well as an
eager transformer.

The complexity in this CL is necessary for aggregate transformers regardless.

[Bob Nystrom, 2014-04-30 19:21:24]

lgtm

https://codereview.chromium.org/255283003/diff/1/pkg/barback/test/package_gra...
File pkg/barback/test/package_graph/declaring_transformer_test.dart (right):

https://codereview.chromium.org/255283003/diff/1/pkg/barback/test/package_gra...
pkg/barback/test/package_graph/declaring_transformer_test.dart:224: test("still
runs apply", () {
On 2014/04/30 00:36:57, nweiz wrote:
> On 2014/04/29 23:46:22, Bob Nystrom wrote:
> > This seems like a perverse edge case to add complexity to support. If a
> > transformer fails when declaring its outputs, why apply it?
> 
> Why not apply it? As far as we know, it'll still function perfectly well as an
> eager transformer.

Maybe, but it's also likely to be broken if it couldn't even declare its
outputs. The main reason I'd say we shouldn't support it would be just to keep
things simpler.

> The complexity in this CL is necessary for aggregate transformers regardless.

Ah, that's fine then.

[nweiz, 2014-05-01 00:48:42]

Committed patchset #1 manually as r35606 (presubmit successful).