Remove NT_PLACE, it's not really used.

Remove NT_PLACE, it's not really used.

operator new(size_t size, void* buffer, sandbox::AllocationType type) just
returns buffer, which is what the default placement new operator does too.

No intended behavior change.

(The motivation is to get rid of the two "return NULL"s in
`operator delete(void* memory, sandbox::AllocationType type)`, as they are
undefined behavior.)

BUG=none

Committed: https://crrev.com/7db050e5a7f0580b0ad7170e0e51c81ddf5079fc
Cr-Commit-Position: refs/heads/master@{#294119}

[Nico, 2014-09-05 21:14:40]

thakis@chromium.org changed reviewers:
+ cpu@chromium.org

[rvargas (doing something else), 2014-09-05 21:41:16]

On 2014/09/05 21:14:41, Nico (hiding) wrote:

I believe the reason for NT_PLACE is that it is used from places that have no
dependencies on the CRT being present (or initialized).

[Nico, 2014-09-08 18:14:56]

On 2014/09/05 21:41:16, rvargas wrote:
> On 2014/09/05 21:14:41, Nico (hiding) wrote:
> 
> I believe the reason for NT_PLACE is that it is used from places that have no
> dependencies on the CRT being present (or initialized).

Oh, that's a good point, I didn't know that. But placement operator new is
defined inline, and it's literally just `inline void* operator now(size_t, void*
p) { return p; }`, so it should get inlined and not add a dependency on the crt.

[Nico, 2014-09-09 15:06:10]

ping

[cpu_(ooo_6.6-7.5), 2014-09-10 00:43:07]

cpu@chromium.org changed reviewers:
+ rvargas@chromium.org
- cpu@chromium.org

[Nico, 2014-09-10 01:06:53]

CPU: you replied with an empty message, was that intentional?
On Sep 9, 2014 5:43 PM, <cpu@chromium.org> wrote:

> https://codereview.chromium.org/544333002/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[rvargas (doing something else), 2014-09-10 01:21:17]

On 2014/09/08 18:14:56, Nico (hiding) wrote:
> On 2014/09/05 21:41:16, rvargas wrote:
> > On 2014/09/05 21:14:41, Nico (hiding) wrote:
> > 
> > I believe the reason for NT_PLACE is that it is used from places that have
no
> > dependencies on the CRT being present (or initialized).
> 
> Oh, that's a good point, I didn't know that. But placement operator new is
> defined inline, and it's literally just `inline void* operator now(size_t,
void*
> p) { return p; }`, so it should get inlined and not add a dependency on the
crt.

OK. I think the assumption at that time was that it was better not to depend on
the actual implementation (many compiler versions ago). I probably never checked
if it was inlined or not.

Shouldn't we have a comment somewhere saying that we expect the placement new to
be inlined, or that we include <new> only for that? I'm thinking about
preventing future changes that may cause a hard dependency to be added without
noticing. That is to say, lgtm.

[Nico, 2014-09-10 05:02:03]

Added a comment.

[Nico, 2014-09-10 05:02:25]

The CQ bit was checked by thakis@chromium.org

[commit-bot: I haz the power, 2014-09-10 05:03:16]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/thakis@chromium.org/544333002/40001

[commit-bot: I haz the power, 2014-09-10 06:14:26]

Committed patchset #3 (id:40001) as b75fae7fdd179e993621928d606d0f8ba4e6e496

[commit-bot: I haz the power, 2014-09-10 06:17:48]

Patchset 3 (id:??) landed as
https://crrev.com/7db050e5a7f0580b0ad7170e0e51c81ddf5079fc
Cr-Commit-Position: refs/heads/master@{#294119}