- Fix build break on Mac OS X by using the proper formatting for pointers....

- Fix build break on Mac OS X by using the proper formatting for pointers
  by working around slightly "wrong" definition of uintptr_t on Mac OS X.
  Verified that this works on both ia32 and x64 on Linux and Mac OS X.


Committed: http://code.google.com/p/v8/source/detail?r=1938

[Erik Corry, 2009-05-12 22:22:23]

LGTM

[Dean McNamee, 2009-05-13 08:24:21]

http://codereview.chromium.org/115252/diff/1007/1009
File src/globals.h (right):

http://codereview.chromium.org/115252/diff/1007/1009#newcode95
Line 95: #define V8PRIxPTR "lx"
We need Id also, it should be ld.  We will fix this up today.

http://codereview.chromium.org/115252/diff/1007/1008
File src/platform-macos.cc (right):

http://codereview.chromium.org/115252/diff/1007/1008#newcode501
Line 501: #error Unsupported Mac OS X host architecture.
I have come around on this, and I like it now.

[iposva, 2009-05-13 16:37:08]

Addressed review comments.

[Dean McNamee, 2009-05-13 16:40:59]

Ok

http://codereview.chromium.org/115252/diff/1014/10
File src/globals.h (right):

http://codereview.chromium.org/115252/diff/1014/10#newcode92
Line 92: // Fix for Mac OS X defining uintptr_t as "unsigned long":
I would probably structure this as an if/else with the defines above (so you
don't have to undef), and I wouldn't say Fix, just say that OS X defines it as
unsigned long.

[iposva, 2009-05-13 16:54:50]

http://codereview.chromium.org/115252/diff/1014/10
File src/globals.h (right):

http://codereview.chromium.org/115252/diff/1014/10#newcode92
Line 92: // Fix for Mac OS X defining uintptr_t as "unsigned long":
On 2009/05/13 16:40:59, Dean McNamee wrote:
> I would probably structure this as an if/else with the defines above (so you
> don't have to undef), and I wouldn't say Fix, just say that OS X defines it as
> unsigned long.

I sort of like your offline suggestion better about reusing the standard
definitions of PRI[x|d]PTR  from inttypes.h in the definition of V8PRI[x|d]PTR.
Unfortunately we are using V8PRIdPTR to print pointer differences (32-bit type
int) and not pointers (32-bit type long int) which is what PRIdPTR expects on
Mac OS X. So that solution would also need an exception.

I am standing by "Fix" for the description. Also in my opinion it is more
readable this way because it clearly marks it as an exception instead of being
buried in a series of nested #if.

P.S. We should all be happy that intptr_t is not any longer typedef'd to "int*"
as it used to in early versions of Mac OS X...

[Dean McNamee, 2009-05-13 17:00:34]

Ok, we should try to clean up the intptr_t/ptrdiff_t etc situation, but
unbreaking the build for now is obviously a good thing to do.

Thanks