64-bit support for Mac OS X in base

64-bit support for Mac OS X in base.

BUG=44127, 18323
TEST=64-bit base should all (mostly) build
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=79069

[Mark Mentovai, 2011-03-22 19:06:35]

Note that there are still a couple of bits of NSS that don’t work. One of them
actually seems to have uncovered a real bug in that I think the SHA-512
implementation we’re building (but not using or testing anywhere) is likely
broken on the Mac. I’ll talk to Wan-Teh and come back around to fix these
tomorrow.

[Robert Sesek, 2011-03-22 19:17:13]

LGTM

On 2011/03/22 19:06:35, Mark Mentovai wrote:
> Note that there are still a couple of bits of NSS that don’t work. One of them
> actually seems to have uncovered a real bug in that I think the SHA-512
> implementation we’re building (but not using or testing anywhere) is likely
> broken on the Mac. I’ll talk to Wan-Teh and come back around to fix these
> tomorrow.

If we're neither using nor testing it, can we stop building it?

[Mark Mentovai, 2011-03-22 19:18:36]

I aim to stop building it, but wtc’s not available right now.

[wtc, 2011-03-22 23:46:33]

LGTM.

The only SHA-512 bug I know of is
https://bugzilla.mozilla.org/show_bug.cgi?id=610162

Please tell me more about the bug you found.

http://codereview.chromium.org/6708096/diff/3001/base/basictypes.h
File base/basictypes.h (right):

http://codereview.chromium.org/6708096/diff/3001/base/basictypes.h#newcode35
base/basictypes.h:35: #if __LP64__ && !defined(OS_MACOSX)
Nit: it's better to test
    defined(__LP64__)
instead of
    __LP64__
here and at line 57 below.

Your changes match the following ifdef in NSPR's prtypes.h:
#if PR_BYTES_PER_LONG == 8 && !defined(__APPLE__)
typedef long PRInt64;
typedef unsigned long PRUint64;

http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/nsprpub/pr/include/prtype...

The best way to fix this mess is to compile Chromium code
that includes any NSPR/NSS public headers with -DNO_NSPR_10_SUPPORT.

Do you want to give that a try?

[Mark Mentovai, 2011-03-22 23:58:23]

The SHA-512 bug was a false alarm, although I still think we should disable
building that for Chrome since we don’t seem to be using it or testing it
anywhere.

I’m fixing prtypes.h in http://codereview.chromium.org/6722011, making it
through the tryservers now.

[wtc, 2011-03-23 00:10:37]

mark: the SHA-512 code is commented out using #if 0.  I use the #if 0
method to make it easier to merge with new versions of the file from
NSS upstream.

[Mark Mentovai, 2011-03-23 00:30:00]

wtc@chromium.org wrote:
> mark: the SHA-512 code is commented out using #if 0.  I use the #if 0
> method to make it easier to merge with new versions of the file from
> NSS upstream.

Excellent. I plan on taking a closer look tomorrow. Thanks for your comments.