Callback support for unbound reference and array arguments.

Callback support for unbound reference and array arguments.

Because the callback object uses const An& for the type of the Run() function in argument forwarding, the code breaks for An=T& or An=T[].  This CL adds in code to modify the parameter type to remove duplicate references, and other fun.

BUG=35223
TEST=new unittests

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=79239

[awong, 2011-03-22 05:09:27]

Fred, your turn!

James, I think this will fix your issue with the reference parameter.

[akalin, 2011-03-24 00:40:19]

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h
File base/callback_internal.h (right):

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h#new...
base/callback_internal.h:83: // This is a typetraits object that's used to take
an argumen types, and
an argumen -> argument
storing, and forwarding -> storing and forwarding

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h#new...
base/callback_internal.h:122: template <typename T>
in this context, and in arguments, T[] means T*, right?

[awong, 2011-03-24 01:37:17]

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h
File base/callback_internal.h (right):

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h#new...
base/callback_internal.h:83: // This is a typetraits object that's used to take
an argumen types, and
On 2011/03/24 00:40:19, akalin wrote:
> an argumen -> argument

Done

> storing, and forwarding -> storing and forwarding

Both you and Will apparently belong to the "no-to-serial-commas" camp. :P

Re-styled.

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h#new...
base/callback_internal.h:122: template <typename T>
On 2011/03/24 00:40:19, akalin wrote:
> in this context, and in arguments, T[] means T*, right?

Yes. As far as I know.  See the comment above for why we adorn a "const."

[akalin, 2011-03-24 01:49:17]

LGTM

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h
File base/callback_internal.h (right):

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h#new...
base/callback_internal.h:122: template <typename T>
On 2011/03/24 01:37:17, awong wrote:
> On 2011/03/24 00:40:19, akalin wrote:
> > in this context, and in arguments, T[] means T*, right?
> 
> Yes. As far as I know.  See the comment above for why we adorn a "const."

It might be less confusing to have the template type as "T*", and/or add a
comment reminding about the T*/T[] equivalence and when it applies.

[awong, 2011-03-24 01:58:56]

Thanks for the review!

Submitting.

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h
File base/callback_internal.h (right):

http://codereview.chromium.org/6718021/diff/4002/base/callback_internal.h#new...
base/callback_internal.h:122: template <typename T>
On 2011/03/24 01:49:18, akalin wrote:
> On 2011/03/24 01:37:17, awong wrote:
> > On 2011/03/24 00:40:19, akalin wrote:
> > > in this context, and in arguments, T[] means T*, right?
> > 
> > Yes. As far as I know.  See the comment above for why we adorn a "const."
> 
> It might be less confusing to have the template type as "T*", and/or add a
> comment reminding about the T*/T[] equivalence and when it applies.

Unforutnately, during the actual pattern match for the type of the template, T[]
is still actually T[] and distinct from T*.  It's just that it effectively
"promotes" to T* when you use it as a parameter.

This code is used to make the template system do the equivalent promotion.