IDL: Remove last traces of [TypeChecking=Unrestricted]

IDL: Remove last traces of [TypeChecking=Unrestricted]

- Remove the Unrestricted value from IDLExtendedAttributes.txt.

- Remove [TypeChecking=Unrestricted] from InternalSettingsGenerated.idl
  which was previously missed since it's a generated file.

- To simplify the code, revert some of the changes made in
    https://codereview.chromium.org/938403007/
  that are no longer needed since TypeChecking now only supports a single
  value (Interface).

BUG=354298
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=191178

[Jens Widell, 2015-03-02 17:29:22]

jl@opera.com changed reviewers:
+ haraken@chromium.org

[Jens Widell, 2015-03-02 17:29:23]

PTAL

[haraken, 2015-03-03 01:23:31]

LGTM

https://codereview.chromium.org/971783002/diff/1/Source/bindings/IDLExtendedA...
File Source/bindings/IDLExtendedAttributes.txt (right):

https://codereview.chromium.org/971783002/diff/1/Source/bindings/IDLExtendedA...
Source/bindings/IDLExtendedAttributes.txt:100: TypeChecking=Interface

Can we change [TypeChecking=Interface] to [TypeChecking], simply?

BTW, how much work would be needed to make [TypeChecking=Interface] the default
and deprecate [TypeChecking=Interface]?

[Jens Widell, 2015-03-03 07:03:35]

On 2015/03/03 01:23:31, haraken wrote:
> LGTM
> 
>
https://codereview.chromium.org/971783002/diff/1/Source/bindings/IDLExtendedA...
> File Source/bindings/IDLExtendedAttributes.txt (right):
> 
>
https://codereview.chromium.org/971783002/diff/1/Source/bindings/IDLExtendedA...
> Source/bindings/IDLExtendedAttributes.txt:100: TypeChecking=Interface
> 
> Can we change [TypeChecking=Interface] to [TypeChecking], simply?

We could, but that would of course involve changing all the IDL files that
actually use it. (There are 224 uses in 110 .idl files.)

Since we're going to get rid of it entirely, one way or another, I'd prefer to
focus effort on that, instead of a simple rename.

> BTW, how much work would be needed to make [TypeChecking=Interface] the
default
> and deprecate [TypeChecking=Interface]?

I haven't really investigated the affected APIs to tell how controversial it
would be just enable type checking for them. The problem with the legacy
behavior is that it makes the argument sort of pseudo-optional/pseudo-nullable,
so you'll need to investigate whether an attribute or argument should actually
be optional/nullable, and of course evaluate the compatibility risk of adding
exception throwing where there previously was none.

But a mostly mechanical initial approach would be:

1) Add [LegacyInterfaceTypeChecking] (or something like that) with the opposite
meaning, and add that to all affected attributes/methods. (No CG changes.)

2) Remove all [TypeChecking=Interface] uses and remove support for the
attribute. (No CG changes.)

That approach gives us explicit flagging of all "wrong" attributes/methods, and
also the correct default behavior, so is a marked improvement, IMO. And, since
there's no CG changes, quite uncontroversial.

[Jens Widell, 2015-03-03 07:05:08]

The CQ bit was checked by jl@opera.com

[commit-bot: I haz the power, 2015-03-03 07:05:56]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/971783002/1

[haraken, 2015-03-03 08:23:20]

> > BTW, how much work would be needed to make [TypeChecking=Interface] the
> default
> > and deprecate [TypeChecking=Interface]?
> 
> I haven't really investigated the affected APIs to tell how controversial it
> would be just enable type checking for them. The problem with the legacy
> behavior is that it makes the argument sort of
pseudo-optional/pseudo-nullable,
> so you'll need to investigate whether an attribute or argument should actually
> be optional/nullable, and of course evaluate the compatibility risk of adding
> exception throwing where there previously was none.
> 
> But a mostly mechanical initial approach would be:
> 
> 1) Add [LegacyInterfaceTypeChecking] (or something like that) with the
opposite
> meaning, and add that to all affected attributes/methods. (No CG changes.)
> 
> 2) Remove all [TypeChecking=Interface] uses and remove support for the
> attribute. (No CG changes.)
> 
> That approach gives us explicit flagging of all "wrong" attributes/methods,
and
> also the correct default behavior, so is a marked improvement, IMO. And, since
> there's no CG changes, quite uncontroversial.

Sounds like a plan!

[commit-bot: I haz the power, 2015-03-03 08:23:41]

Committed patchset #1 (id:1) as
https://src.chromium.org/viewvc/blink?view=rev&revision=191178