Fix selector name check, and make the interceptor participate in speculatively propagating types.

Fix selector name check, and make the interceptor participate in speculatively propagating types.
Committed: https://code.google.com/p/dart/source/detail?r=17055

[ngeoffray, 2013-01-15 10:30:32]

TBR to unbreak the build.

[ngeoffray, 2013-01-15 10:32:27]

https://codereview.chromium.org/11885039/diff/1/sdk/lib/_internal/compiler/im...
File sdk/lib/_internal/compiler/implementation/ssa/nodes.dart (right):

https://codereview.chromium.org/11885039/diff/1/sdk/lib/_internal/compiler/im...
sdk/lib/_internal/compiler/implementation/ssa/nodes.dart:2408: Compiler
compiler) {
This was the main reason for the unit test to fail: the unit test was looking
for a type guard, but since the interceptor did not say the type it wanted, the
type guard was not inserted. Before, a "-a" in Dart code that translates to the
SSA HNegate was always saying it wanted a number as input.

[ahe, 2013-01-15 10:39:06]

This exceeds my comfort-level for a TBR submit.

[ngeoffray, 2013-01-15 10:46:34]

On 2013/01/15 10:39:06, ahe wrote:
> This exceeds my comfort-level for a TBR submit.

Which part of the CL do you think is controversial?
FYI, the main change is the interceptor now saying what it wants as input. This
is in order to partially have the same behavior as before.

[ahe, 2013-01-15 11:06:05]

On 2013/01/15 10:46:34, ngeoffray wrote:
> On 2013/01/15 10:39:06, ahe wrote:
> > This exceeds my comfort-level for a TBR submit.
> 
> Which part of the CL do you think is controversial?
> FYI, the main change is the interceptor now saying what it wants as input.
This
> is in order to partially have the same behavior as before.

I don't think there's anything controversial in this CL. However, it is not a
trivial change, so I would have approached this by reverting the CL which broke
the build.

BTW: it looks like this CL didn't fix all problems.

[ngeoffray, 2013-01-15 11:25:37]

On 2013/01/15 11:06:05, ahe wrote:
> On 2013/01/15 10:46:34, ngeoffray wrote:
> > On 2013/01/15 10:39:06, ahe wrote:
> > > This exceeds my comfort-level for a TBR submit.
> > 
> > Which part of the CL do you think is controversial?
> > FYI, the main change is the interceptor now saying what it wants as input.
> This
> > is in order to partially have the same behavior as before.
> 
> I don't think there's anything controversial in this CL. However, it is not a
> trivial change, so I would have approached this by reverting the CL which
broke
> the build.

OK, thanks for letting me know. I'll reconsider my threshold for TBR the next
times.

> 
> BTW: it looks like this CL didn't fix all problems.

Yes, CL sent: https://codereview.chromium.org/11876039/.