Fix issue 18435 (2nd attempt).

Fix issue 18435 (2nd attempt).
Add regression test.

R=hausner@google.com

Committed: https://code.google.com/p/dart/source/detail?r=36402

[regis, 2014-05-20 19:09:29]

I realize that this solution is too brittle.
I notice now that the parser is accessing the variable at position 0 in the
scope, assuming it is the receiver.

Either we add a ParameterAt(pos) that skips aliases (as in this cl), or add an
AddParameter() that inserts parameter variables at the beginning of the list
(requires fewer changes). Adding both would actually be a good thing.

Let me know what you think.

Thanks,
Regis

[hausner, 2014-05-20 20:14:23]

Regis, I agree that this solution is probably not the right thing. As you
mention, the parser assumes in some places that the first entry in the scope is
the receiver. If that assumption is wrong, we probably fail silently, which
would be a regression from the current bug.

How about inserting the parameters in front of any aliases that may be in the
scope? Or how about opening a new scope for the parameters? (Not sure whether
the latter would break other assumptions, e.g. that the outermost scope contains
the parameters.)

[regis, 2014-05-20 23:16:15]

On 2014/05/20 20:14:23, hausner wrote:
> Regis, I agree that this solution is probably not the right thing. As you
> mention, the parser assumes in some places that the first entry in the scope
is
> the receiver. If that assumption is wrong, we probably fail silently, which
> would be a regression from the current bug.
> 
> How about inserting the parameters in front of any aliases that may be in the
> scope? Or how about opening a new scope for the parameters? (Not sure whether
> the latter would break other assumptions, e.g. that the outermost scope
contains
> the parameters.)

Thanks Matthias.

I added an InsertParameterAt(pos, param) which is a safer and simpler change.

PTAL

[hausner, 2014-05-21 00:14:23]

LGTM.

[regis, 2014-05-21 00:32:37]

Committed patchset #2 manually as r36402 (presubmit successful).

[regis, 2014-05-21 00:32:46]

Thanks!