Workaround dubious casting between CXFA_Object and void* in FXJSE

Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening
when setting a CXFA_ThisProxy (Derived, virtual) to be a global
object (void*). This would then be cast back to a CFXA_Object
(Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
Committed: https://pdfium.googlesource.com/pdfium/+/3a005f22703b9303a306bf34cbd17c3729f763aa

[Tom Sepez, 2016-05-27 20:14:05]

Description was changed from

==========
Fix type confusion in FXJSE to V8 bindings

BUG=
==========

to

==========
Fix type confusion in FXJSE to V8 bindings

BUG=613607
==========

[Tom Sepez, 2016-05-27 21:41:46]

Description was changed from

==========
Fix type confusion in FXJSE to V8 bindings

BUG=613607
==========

to

==========
Cleanup some arguments and functions in FXJSE

BUG=613607
==========

[Tom Sepez, 2016-05-28 00:18:26]

Description was changed from

==========
Cleanup some arguments and functions in FXJSE

BUG=613607
==========

to

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice to C-style
reinterpret a Derived* ptr to void* and then back to a Base* ptr.  One place
where this blows up is
if Derived has virtual functions but Base does not, in which case the world will
be offst by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening when setting
a CXFA_ThisProxy to be a global object.  FXJSE global objects need to be made
type-safe.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

[Tom Sepez, 2016-05-28 00:18:26]

tsepez@chromium.org changed reviewers:
+ thestig@chromium.org

[Tom Sepez, 2016-05-28 00:20:11]

Description was changed from

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice to C-style
reinterpret a Derived* ptr to void* and then back to a Base* ptr.  One place
where this blows up is
if Derived has virtual functions but Base does not, in which case the world will
be offst by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening when setting
a CXFA_ThisProxy to be a global object.  FXJSE global objects need to be made
type-safe.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

to

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is
if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening when setting
a CXFA_ThisProxy (Derived, virtual) to be a global object. This would then be
cast back to a CFXA_Object (Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

[Tom Sepez, 2016-05-28 00:21:42]

Lei, for review.  I wanted to get something in before we all left.  I may not
get back to this for some time.  Sorry the fix is so kludgy.

[Tom Sepez, 2016-05-28 00:22:17]

Description was changed from

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is
if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening when setting
a CXFA_ThisProxy (Derived, virtual) to be a global object. This would then be
cast back to a CFXA_Object (Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

to

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is
if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening
when setting a CXFA_ThisProxy (Derived, virtual) to be a global
object (void*). This would then be cast back to a CFXA_Object
(Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

[Tom Sepez, 2016-05-28 00:24:06]

Description was changed from

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is
if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening
when setting a CXFA_ThisProxy (Derived, virtual) to be a global
object (void*). This would then be cast back to a CFXA_Object
(Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

to

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening
when setting a CXFA_ThisProxy (Derived, virtual) to be a global
object (void*). This would then be cast back to a CFXA_Object
(Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

[Lei Zhang, 2016-05-28 00:44:39]

The CQ bit was checked by thestig@chromium.org

[Lei Zhang, 2016-05-28 00:44:40]

lgtm

[commit-bot: I haz the power, 2016-05-28 00:44:46]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2015143005/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2015143005/60001

[commit-bot: I haz the power, 2016-05-28 00:45:07]

Description was changed from

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening
when setting a CXFA_ThisProxy (Derived, virtual) to be a global
object (void*). This would then be cast back to a CFXA_Object
(Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607
==========

to

==========
Workaround dubious casting between CXFA_Object and void* in FXJSE

This is just a crock to get things working until we fix the
underlying issue.

When there's single-inheritance, it may often work in practice
to C-style (reinterpret) cast a Derived* ptr to void* and then
back to a Base* ptr.  One place where this blows up is if 
Derived has virtual functions but Base does not, in which case
the world will be offset by the size of a vtable ptr.

Because of the use of void* types in FXJSE, the above was happening
when setting a CXFA_ThisProxy (Derived, virtual) to be a global
object (void*). This would then be cast back to a CFXA_Object
(Base, non-virtual) and chaos is ensured.

Not sure how far back this goes.

Along the way, pick up some tidying which was necessary for
simplicity while tracking this down.

BUG=613607

Committed:
https://pdfium.googlesource.com/pdfium/+/3a005f22703b9303a306bf34cbd17c3729f7...
==========

[commit-bot: I haz the power, 2016-05-28 00:45:08]

Committed patchset #4 (id:60001) as
https://pdfium.googlesource.com/pdfium/+/3a005f22703b9303a306bf34cbd17c3729f7...