Fix definition of CFAllocator (malloc_zone_t changed size between 10.5 and 10...

Fix definition of CFAllocator (malloc_zone_t changed size between 10.5 and 10.6).

BUG=none
TEST=no visible change, works when compiled against 10.6 SDK


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=47676

[Avi (use Gerrit), 2010-05-18 17:16:59]

Hey, good thing we were compiling against the 10.5 SDK, right? :)

[Mark Mentovai, 2010-05-19 15:59:53]

http://codereview.chromium.org/2131010/diff/1/2
File base/process_util_mac.mm (right):

http://codereview.chromium.org/2131010/diff/1/2#newcode424
base/process_util_mac.mm:424: void* cf_runtime_base_1;
Not 64-bit-friendly. You should provide a definition of ChromeCFRuntimeBase
mimicing __CFRuntimeBase from CFRuntime.h.

http://codereview.chromium.org/2131010/diff/1/2#newcode426
base/process_util_mac.mm:426: void* size_callback;
It’d be nice if you borrowed the actual definitions with types and arguments
instead of having a whole bunch of void pointers.

http://codereview.chromium.org/2131010/diff/1/2#newcode436
base/process_util_mac.mm:436: struct malloc_introspection_t *introspect;
This is the only one where the * is attached to the field name and not the type.
Be consistent.

http://codereview.chromium.org/2131010/diff/1/2#newcode437
base/process_util_mac.mm:437: void* version;
reserved5?

[Avi (use Gerrit), 2010-05-19 16:16:15]

http://codereview.chromium.org/2131010/diff/1/2
File base/process_util_mac.mm (right):

http://codereview.chromium.org/2131010/diff/1/2#newcode424
base/process_util_mac.mm:424: void* cf_runtime_base_1;
On 2010/05/19 15:59:53, Mark Mentovai wrote:
> Not 64-bit-friendly. You should provide a definition of ChromeCFRuntimeBase
> mimicing __CFRuntimeBase from CFRuntime.h.

OK, but it is 64-bit friendly. CFRuntimeBase is always two pointer sizes big.

http://codereview.chromium.org/2131010/diff/1/2#newcode426
base/process_util_mac.mm:426: void* size_callback;
On 2010/05/19 15:59:53, Mark Mentovai wrote:
> It’d be nice if you borrowed the actual definitions with types and arguments
> instead of having a whole bunch of void pointers.

Why? We're not using them, and they're a compatibility shim on Apple's side
anyway.

http://codereview.chromium.org/2131010/diff/1/2#newcode436
base/process_util_mac.mm:436: struct malloc_introspection_t *introspect;
On 2010/05/19 15:59:53, Mark Mentovai wrote:
> This is the only one where the * is attached to the field name and not the
type.
> Be consistent.

Done.

http://codereview.chromium.org/2131010/diff/1/2#newcode437
base/process_util_mac.mm:437: void* version;
On 2010/05/19 15:59:53, Mark Mentovai wrote:
> reserved5?

In the correspondence between the allocator and the malloc_zone_t, this field is
the version field. OTOH, in the allocator it's a pointer and the zone it's an
int, so we already have a mismatch...

[Mark Mentovai, 2010-05-19 16:25:37]

LGTM (I’d still prefer real definitions instead of a bunch of void*, but
whatever.)

[Avi (use Gerrit), 2010-05-19 16:36:10]

:)

[Mark Mentovai, 2010-05-19 16:37:25]

Much tastier.