fix x32 handling of Atomic64

fix x32 handling of Atomic64

The x32 logic for the size of Atomic64 handles NaCL, but misses
the Linux case.  Check the standard __ILP32__ to handle that too.

This has been fixed in the Chromium base tree already:
https://codereview.chromium.org/12186005/

BUG=chromium-os:36866
TEST=compiled the code for x86_64 (64bit) & x86_64 (x32)

Committed: http://code.google.com/p/v8/source/detail?r=13824

[Brad Chen (chromium), 2013-03-04 21:53:16]

On 2013/03/02 09:50:35, vapier wrote:

The change looks fine, although I don't quite understand the review process.
Does the V8 project have a notion of committers and owners and such? Or can any
Chromium committer land changes?

[vapier, 2013-03-04 22:32:37]

sorry, i thought i had added more reviewers here.  the v8 guys have their own
set of reviewers last i merged stuff.

[Brad Chen (chromium), 2013-03-04 23:27:32]

Looking at this more carefully...

https://codereview.chromium.org/12374065/diff/1/src/atomicops.h
File src/atomicops.h (right):

https://codereview.chromium.org/12374065/diff/1/src/atomicops.h#newcode61
src/atomicops.h:61: #if defined(__ILP32__) || defined(__APPLE__)
So, actually trying this in my own build, I'm not sure we actually want
V8_HOST_ARCH_64_BIT for the ILP32 build. Have you thought through this?

Note for example in globals.h:247; this code would have to be fixed too.

[vapier, 2013-03-04 23:46:21]

i have no opinion on that matter :).  this is a common atomicops header though
and this fix has been landed in the other places.  it certainly doesn't lock us
into a decision either way when it comes to V8_HOST_ARCH_64_BIT vs
V8_HOST_ARCH_32_BIT.  this code path is unreachable in the latter case, and
necessary in the former.

[danno, 2013-03-05 12:59:12]

Hi Sven, can you please take a look at this and shepherd it through to land?

[Sven Panne, 2013-03-05 13:51:49]

lgtm

[Sven Panne, 2013-03-05 13:54:08]

Committed patchset #1 manually as r13824 (presubmit successful).