Adds base::WrapFunction to wrap lambdas as base::Callbacks

Adds base::WrapFunction to wrap lambdas as base::Callbacks

BUG=

[Ken Rockot(use gerrit already), 2016-06-22 22:58:11]

rockot@chromium.org changed reviewers:
+ dcheng@chromium.org, yzshen@chromium.org

[Ken Rockot(use gerrit already), 2016-06-22 22:58:12]

So I don't really anticipate style or base OWNERS being OK with this.

However, lambdas made a lot of our Mojo test code easier to read and write IMHO.
Yuzhu suggested I expand on a utility built for one of these sets of tests
(converted here) to make it more generally useful. So that's what this is. Note
that there are many other tests which could benefit from this helper in lieu of
defining several one-off, out-of-line functions to be bound.

I thought it was worth pitching something like this to at least provoke a
discussion. WDYT?

[yzshen1, 2016-06-22 23:05:54]

I like this because it really makes our tests much easier to read!

If the owners don't feel like to approve this (and I understand they have good
reasons to be cautious...), maybe we can make this a testing utility?

https://codereview.chromium.org/2090013004/diff/1/base/function_util.h
File base/function_util.h (right):

https://codereview.chromium.org/2090013004/diff/1/base/function_util.h#newcode19
base/function_util.h:19: static Result Run(FuncType func, Args... args) {
Do we need Args&&... (here and below)?

[Ken Rockot(use gerrit already), 2016-06-22 23:14:43]

Providing this as a test-only utility could be a decent compromise.

https://codereview.chromium.org/2090013004/diff/1/base/function_util.h
File base/function_util.h (right):

https://codereview.chromium.org/2090013004/diff/1/base/function_util.h#newcode19
base/function_util.h:19: static Result Run(FuncType func, Args... args) {
On 2016/06/22 at 23:05:54, yzshen1 wrote:
> Do we need Args&&... (here and below)?

This wouldn't work - we can't for example base::Bind a function with a signature
void(Foo&&) and get a base::Callback<void(Foo)>.

[dcheng, 2016-06-23 19:23:01]

dcheng@chromium.org changed reviewers:
+ danakj@chromium.org
- yzshen@chromium.org

[dcheng, 2016-06-23 19:23:01]

I think I'd be a lot more comfortable with this if we disabled this wholly on
capturing lambdas. In theory, we wouldn't need a wrapping helper then, since
non-capturing lambdas can be converted to function pointers, right?

[dcheng, 2016-06-23 19:23:29]

Description was changed from

==========
Adds base::WrapFunction to wrap lambdas as base::Callbacks

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

to

==========
Adds base::WrapFunction to wrap lambdas as base::Callbacks

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

[dcheng, 2016-06-23 19:23:29]

dcheng@chromium.org changed reviewers:
+ yzshen@chromium.org
- danakj@chromium.org

[Ken Rockot(use gerrit already), 2016-06-23 19:35:36]

On 2016/06/23 at 19:23:01, dcheng wrote:
> I think I'd be a lot more comfortable with this if we disabled this wholly on
capturing lambdas. In theory, we wouldn't need a wrapping helper then, since
non-capturing lambdas can be converted to function pointers, right?

Unfortunately the usefulness of this goes down quite a bit if we completely
disallow capturing. I don't think it would be worth adding at all with that
restriction. In the tests updated by this CL for example, every use of the
wrapper requires a capture in order to be useful.

Having said that, I can understand why we wouldn't want to add something like
this, it's just kind of a shame.

[dcheng, 2016-06-23 19:37:57]

On 2016/06/23 19:35:36, Ken Rockot wrote:
> On 2016/06/23 at 19:23:01, dcheng wrote:
> > I think I'd be a lot more comfortable with this if we disabled this wholly
on
> capturing lambdas. In theory, we wouldn't need a wrapping helper then, since
> non-capturing lambdas can be converted to function pointers, right?
> 
> Unfortunately the usefulness of this goes down quite a bit if we completely
> disallow capturing. I don't think it would be worth adding at all with that
> restriction. In the tests updated by this CL for example, every use of the
> wrapper requires a capture in order to be useful.
> 
> Having said that, I can understand why we wouldn't want to add something like
> this, it's just kind of a shame.

If the code wants to:
- capture things
- use lambdas
- emit a callback type

It'd still work with a non-capturing lambda, you'd just have to use base::Bind()
to capture the arguments. That doesn't seem so bad to me?

[Ken Rockot(use gerrit already), 2016-06-23 19:46:39]

On 2016/06/23 at 19:37:57, dcheng wrote:
> On 2016/06/23 19:35:36, Ken Rockot wrote:
> > On 2016/06/23 at 19:23:01, dcheng wrote:
> > > I think I'd be a lot more comfortable with this if we disabled this wholly
on
> > capturing lambdas. In theory, we wouldn't need a wrapping helper then, since
> > non-capturing lambdas can be converted to function pointers, right?
> > 
> > Unfortunately the usefulness of this goes down quite a bit if we completely
> > disallow capturing. I don't think it would be worth adding at all with that
> > restriction. In the tests updated by this CL for example, every use of the
> > wrapper requires a capture in order to be useful.
> > 
> > Having said that, I can understand why we wouldn't want to add something
like
> > this, it's just kind of a shame.
> 
> If the code wants to:
> - capture things
> - use lambdas
> - emit a callback type
> 
> It'd still work with a non-capturing lambda, you'd just have to use
base::Bind() to capture the arguments. That doesn't seem so bad to me?

Ah, good point. I'll see what that looks like.