Make int an abstract class.

Make int an abstract class.

This looks deceptively simple (the only real change is in init, which
just copies the code for double, and the rest are just renaming for
consistency). It seems to work.

Committed: https://code.google.com/p/dart/source/detail?r=12304

[Lasse Reichstein Nielsen, 2012-09-12 08:07:46]

If it isn't this simple, I'll stop trying :)

[Lasse Reichstein Nielsen, 2012-09-12 11:01:54]

Sorry about the spam uploading. CL was making a fuzz.

[Lasse Reichstein Nielsen, 2012-09-12 11:03:24]

https://chromiumcodereview.appspot.com/10933039/diff/5001/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://chromiumcodereview.appspot.com/10933039/diff/5001/runtime/vm/object.c...
runtime/vm/object.cc:2735: RawType* Type::IntType() {
Why is it Type::BoolType but Type::Double? Should it be Int or IntType?

[srdjan, 2012-09-12 11:59:47]

LGTM from my part, but I'd like Matthias and Regis (added him to the list of
reviewers) to have a look at it.

[hausner, 2012-09-12 15:55:42]

LGTM 2

I feared int would be harder to convert. Nice job!

[regis, 2012-09-12 16:30:25]

LGTM

A general comment, not specific to this cl:
It would be nice to use the same naming conventions for int, double, and bool.
We use integer_implementation_class, but double_class. I would expect we need to
restrict implementing the bool interface as we do for int and double, but the
code is different and I am not sure why.

Thanks!
Regis

https://chromiumcodereview.appspot.com/10933039/diff/8001/runtime/vm/class_fi...
File runtime/vm/class_finalizer.cc (right):

https://chromiumcodereview.appspot.com/10933039/diff/8001/runtime/vm/class_fi...
runtime/vm/class_finalizer.cc:305: Type::Handle(Type::IntType()).type_class() ==
super_class.raw()) {
I do not quite understand why kIntegerCid (and actually kDoubleCid) are not
sufficient here. I guess they refer to the implementation rather than the
interface, but then, how did it work for int before?
Lasse, it is not fair to ask you those questions, but maybe you can add a TODO
with my name if you cannot answer them. Thanks!

[Lasse Reichstein Nielsen, 2012-09-13 07:19:21]

https://chromiumcodereview.appspot.com/10933039/diff/8001/runtime/vm/class_fi...
File runtime/vm/class_finalizer.cc (right):

https://chromiumcodereview.appspot.com/10933039/diff/8001/runtime/vm/class_fi...
runtime/vm/class_finalizer.cc:305: Type::Handle(Type::IntType()).type_class() ==
super_class.raw()) {
I think it worked for int before because it was an interface, and you can't
extend an interface anyway. I.e., it was caught somewhere else.
I have no idea why this works the way it does (or how it works at all) - I have
been working entirely by pattern-matching.
Adding TODO.

[Lasse Reichstein Nielsen, 2012-09-13 07:38:41]

On 2012/09/12 16:30:25, regis wrote:
> I would expect we need to restrict implementing the bool interface
> as we do for int and double, but the code is different and I am not
> sure why.

True, bool is also a class, and it's prevented from being extended and
implemented too, without this extra line.

You also need to prevent extending and implementing Null. (There should be a
Null class in the core library, which there isn't now for some reason. I'll
probably add it soon and see how much will crash :).



/L