make window interceptors AllCanRead

make window interceptors AllCanRead

R=haraken@chromium.org, jochen@chromium.org

BUG=404300

[dcarney, 2015-02-11 11:49:04]

dcarney@chromium.org changed reviewers:
+ haraken@chromium.org, jochen@chromium.org

[dcarney, 2015-02-11 11:50:29]

ptal

haraken@ please confirm the generated bindings stuff and idl stuff looks
alright.  note that i'm setting up the interceptor flags the way i'm doing so
that i can add several more in upcoming cls

[haraken, 2015-02-11 14:22:46]

I might want to see a more detailed CL description. What does [AllCanRead]
imply?

> note that i'm setting up the interceptor flags the way i'm doing so
that i can add several more in upcoming cls

Does this mean that you're going to use [AllCanRead] in all other indexed/named
properties?

[dcarney, 2015-02-11 14:27:47]

On 2015/02/11 14:22:46, haraken wrote:
> I might want to see a more detailed CL description. What does [AllCanRead]
> imply?

it means exactly the same thing v8::ALL_CAN_READ means, that the property is
readable even if the security check fails.

> Does this mean that you're going to use [AllCanRead] in all other
indexed/named
> properties?

no, i think these are the only ones, but i'll have to check.  i'm going to add
more types of interceptor flags, that do things like skip intercepting existing
properties

[haraken, 2015-02-11 14:33:39]

On 2015/02/11 14:27:47, dcarney wrote:
> On 2015/02/11 14:22:46, haraken wrote:
> > I might want to see a more detailed CL description. What does [AllCanRead]
> > imply?
> 
> it means exactly the same thing v8::ALL_CAN_READ means, that the property is
> readable even if the security check fails.
> 
> > Does this mean that you're going to use [AllCanRead] in all other
> indexed/named
> > properties?
> 
> no, i think these are the only ones, but i'll have to check.  i'm going to add
> more types of interceptor flags, that do things like skip intercepting
existing
> properties

Will the flags be used by Window only? Then you can hard-code 'Window' in the
IDL compiler instead of adding new IDL attributes.

[dcarney, 2015-02-11 14:36:03]

> Will the flags be used by Window only? Then you can hard-code 'Window' in the
> IDL compiler instead of adding new IDL attributes.

the other flags will be used for most interceptors.  It's possible that
AllCanRead may be only used by Window

[Jens Widell, 2015-02-11 14:46:23]

On 2015/02/11 14:36:03, dcarney wrote:
> > Will the flags be used by Window only? Then you can hard-code 'Window' in
the
> > IDL compiler instead of adding new IDL attributes.
> 
> the other flags will be used for most interceptors.  It's possible that
> AllCanRead may be only used by Window

We could also reuse [DoNotCheckSecurity] which has this effect on regular
attributes today.

[dcarney, 2015-02-11 16:52:06]

On 2015/02/11 14:46:23, Jens Widell wrote:
> On 2015/02/11 14:36:03, dcarney wrote:
> > > Will the flags be used by Window only? Then you can hard-code 'Window' in
> the
> > > IDL compiler instead of adding new IDL attributes.
> > 
> > the other flags will be used for most interceptors.  It's possible that
> > AllCanRead may be only used by Window
> 
> We could also reuse [DoNotCheckSecurity] which has this effect on regular
> attributes today.

DoNotCheckSecurity has slightly different semantics, returning different objects
to different contexts, while this returns the same (global) objects to different
contexts.  If Window is the only object in question that will have this, I don't
mind have a specific snippet for that and no attribute.

[Jens Widell, 2015-02-11 17:01:31]

On 2015/02/11 16:52:06, dcarney wrote:
> On 2015/02/11 14:46:23, Jens Widell wrote:
> > On 2015/02/11 14:36:03, dcarney wrote:
> > > > Will the flags be used by Window only? Then you can hard-code 'Window'
in
> > the
> > > > IDL compiler instead of adding new IDL attributes.
> > > 
> > > the other flags will be used for most interceptors.  It's possible that
> > > AllCanRead may be only used by Window
> > 
> > We could also reuse [DoNotCheckSecurity] which has this effect on regular
> > attributes today.
> 
> DoNotCheckSecurity has slightly different semantics, returning different
objects
> to different contexts, while this returns the same (global) objects to
different
> contexts.  If Window is the only object in question that will have this, I
don't
> mind have a specific snippet for that and no attribute.

Really? On attributes, the only effect I can find is that the attribute is
installed with some combination of v8::ALL_CAN_{READ,WRITE} instead of
v8::DEFAULT. (See access_control_list() in v8_attributes.py.)

[dcarney, 2015-02-11 19:12:15]

> Really? On attributes, the only effect I can find is that the attribute is
> installed with some combination of v8::ALL_CAN_{READ,WRITE} instead of
> v8::DEFAULT. (See access_control_list() in v8_attributes.py.)

there's a whole bunch OriginSafeMethod stuff generated for those I think as
well, but i can see your point of view.  I think DoNotCheckSecurity works here
instead of AllCanRead.

[jochen (gone - plz use gerrit), 2015-02-12 09:02:04]

updated to use DoNotCheckSecurity

ptal

[haraken, 2015-02-12 09:05:51]

On 2015/02/12 09:02:04, jochen (slow) wrote:
> updated to use DoNotCheckSecurity
> 
> ptal

Did you upload a new patch?

[jochen (gone - plz use gerrit), 2015-02-12 09:43:00]

meh, it didn't let me upload to this issue

CL is here: https://codereview.chromium.org/921683002

[jochen (gone - plz use gerrit), 2015-02-12 09:43:09]

closing this one