Fix wrong type test optimization when testing against a signature class: we must always consider th…

Fix wrong type test optimization when testing against a signature class: we must always consider the instance type arguments as well. FIxes failing tests when run with --optimization-counter-threshold=5.

R=regis@google.com

Committed: https://code.google.com/p/dart/source/detail?r=24961

[regis, 2013-07-12 00:29:30]

LGTM with comments.

https://codereview.chromium.org/18807007/diff/9001/runtime/lib/object.cc
File runtime/lib/object.cc (right):

https://codereview.chromium.org/18807007/diff/9001/runtime/lib/object.cc#newc...
runtime/lib/object.cc:131: const Bool& result = Bool::Handle(
You could spare the handle creation that you just introduced by using a bool
instead of a Bool for 'result'.

Actually, you should not use 'result' below, but 'is_instance_of' or else you
display the wrong result in case of negate.value().

https://codereview.chromium.org/18807007/diff/9001/runtime/vm/flow_graph_comp...
File runtime/vm/flow_graph_compiler_arm.cc (right):

https://codereview.chromium.org/18807007/diff/9001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler_arm.cc:244: if (!type_class.IsSignatureClass()) {
A comment explaining why we do not do a class check for signature classes could
help in the future.

https://codereview.chromium.org/18807007/diff/9001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler_arm.cc:517: // non-parameterized class or with a
raw dst type of a parameterized class.
Should this comment be updated?

[srdjan, 2013-07-12 16:56:11]

https://codereview.chromium.org/18807007/diff/9001/runtime/lib/object.cc
File runtime/lib/object.cc (right):

https://codereview.chromium.org/18807007/diff/9001/runtime/lib/object.cc#newc...
runtime/lib/object.cc:131: const Bool& result = Bool::Handle(
On 2013/07/12 00:29:31, regis wrote:
> You could spare the handle creation that you just introduced by using a bool
> instead of a Bool for 'result'.
> 
> Actually, you should not use 'result' below, but 'is_instance_of' or else you
> display the wrong result in case of negate.value().

Done.

https://codereview.chromium.org/18807007/diff/9001/runtime/vm/flow_graph_comp...
File runtime/vm/flow_graph_compiler_arm.cc (right):

https://codereview.chromium.org/18807007/diff/9001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler_arm.cc:244: if (!type_class.IsSignatureClass()) {
On 2013/07/12 00:29:31, regis wrote:
> A comment explaining why we do not do a class check for signature classes
could
> help in the future.

Added comment:

Signature class is an instantiated parameterized type.

https://codereview.chromium.org/18807007/diff/9001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler_arm.cc:517: // non-parameterized class or with a
raw dst type of a parameterized class.
On 2013/07/12 00:29:31, regis wrote:
> Should this comment be updated?

Added:
non-signature class

[srdjan, 2013-07-12 17:31:24]

Committed patchset #4 manually as r24961 (presubmit successful).