Add support for Mac OS X 64bit builds with GYP

Add support for Mac OS X 64bit builds with GYP

Note that in order to build for 64bits mode, you'll have
to specify the target architecture explicitely, the default
is still 32bits for Mac OS X.

Example with make and gcc:

$ export GYP_GENERATORS=make
$ make dependencies
$ make -j 8 library=shared x64.release

Example with make and clang:

$ export GYP_GENERATORS=make
$ export CC=/usr/bin/clang
$ export CXX=/usr/bin/clang++
$ export GYP_DEFINES="clang=1"
$ make dependencies
$ make -j 8 library=shared x64.release

Example with xcode:

$ export GYP_GENERATORS=xcode
$ build/gyp_v8 -Dtarget_arch=x64
$ xcodebuild -project build/all.xcodeproj -configuration Release

Contributed by Filipe David Manana <fdmanana@gmail.com>;

BUG=
TEST=

Committed: https://code.google.com/p/v8/source/detail?r=11199

[fdmanana, 2012-03-23 16:26:42]

From the v8-dev thread:

http://groups.google.com/group/v8-dev/browse_thread/thread/5e1aa99b9ab9a6d5

[fdmanana, 2012-03-24 09:43:49]

Patchset 2 removes an unnecessary GYP condition when use_system_v8 != 0

[fdmanana, 2012-03-28 14:29:38]

Fixed a typo in the commit message (explictely -> explicitly) and rebased
against latest bleeding_edge branch.

[Jakob Kummerow, 2012-03-28 16:08:51]

Thanks again for doing this. The patch looks pretty good and works fine
according to my testing so far. I have a few comments below.

The biggest issue I see is that AFAIK the Chromium team wants to be able to
choose the target architecture inside Xcode, and potentially build "fat"
(combined 32/64bit) binaries eventually. I'm not saying you have to do all the
work to make that happen, but it'd be unfortunate if your changes were working
against this goal. Unfortunately, I don't know how architecture selection
support in Xcode works, nor what the current status of the plans in that regard
is.

http://codereview.chromium.org/9808065/diff/5001/Makefile
File Makefile (right):

http://codereview.chromium.org/9808065/diff/5001/Makefile#newcode269
Makefile:269: --revision 1251
Is there a specific reason you picked revision 1251? Chromium as of yesterday
uses r1282; if it's stable enough for Chromium it should be stable enough for
us.

http://codereview.chromium.org/9808065/diff/5001/build/common.gypi
File build/common.gypi (right):

http://codereview.chromium.org/9808065/diff/5001/build/common.gypi#newcode322
build/common.gypi:322: ['OS=="mac" and target_arch=="ia32"', {
Since these definitions (here and below) are independent from whether we build
in Release or Debug mode, how about moving them to a general OS=="mac" section
above? (To be precise, there's a big OS!="mac" section, which could get an
else-branch containing this.)

http://codereview.chromium.org/9808065/diff/5001/build/gyp_v8
File build/gyp_v8 (right):

http://codereview.chromium.org/9808065/diff/5001/build/gyp_v8#newcode3
build/gyp_v8:3: # Copyright 2010 the V8 project authors. All rights reserved.
nit: s/2010/2012/

Also, feel free to add yourself to the AUTHORS file ;-)

http://codereview.chromium.org/9808065/diff/5001/tools/gyp/v8.gyp
File tools/gyp/v8.gyp (right):

http://codereview.chromium.org/9808065/diff/5001/tools/gyp/v8.gyp#newcode63
tools/gyp/v8.gyp:63: 'xcode_settings': {'OTHER_LDFLAGS': ['-dynamiclib',
'-all_load']},
80 cols per line, please.

http://codereview.chromium.org/9808065/diff/5001/tools/gyp/v8.gyp#newcode542
tools/gyp/v8.gyp:542: ['v8_target_arch=="ia32"', {
From crbug.com/44205, I gather that it should be possible to choose the
architecture from within Xcode; v8_target_arch=="mac" (or target_arch=="mac", I
don't care) seems to be how the desire to have this option is indicated. Can you
make that work?
Alternatively, would you mind leaving the 'v8_target_arch=="mac"' condition in
so that any progress in that area is not hampered?

[fdmanana, 2012-03-28 16:49:29]

Thanks Jakob. Your comments helped me getting some more insight into the build
system/goals (it's my first contact with GYP). They also make the patch set
smaller and more clear.
Uploading new patch set.

http://codereview.chromium.org/9808065/diff/5001/Makefile
File Makefile (right):

http://codereview.chromium.org/9808065/diff/5001/Makefile#newcode269
Makefile:269: --revision 1251
On 2012/03/28 16:08:51, Jakob wrote:
> Is there a specific reason you picked revision 1251? Chromium as of yesterday
> uses r1282; if it's stable enough for Chromium it should be stable enough for
> us.
Well, 1026 didn't work, so at the time I picked the latest GYP revision that
worked, which was 1251. I will update to 1282 if it works as well.

http://codereview.chromium.org/9808065/diff/5001/build/gyp_v8
File build/gyp_v8 (right):

http://codereview.chromium.org/9808065/diff/5001/build/gyp_v8#newcode3
build/gyp_v8:3: # Copyright 2010 the V8 project authors. All rights reserved.
On 2012/03/28 16:08:51, Jakob wrote:
> nit: s/2010/2012/
> 
> Also, feel free to add yourself to the AUTHORS file ;-)
Thank you, that's very kind :)

http://codereview.chromium.org/9808065/diff/5001/tools/gyp/v8.gyp
File tools/gyp/v8.gyp (right):

http://codereview.chromium.org/9808065/diff/5001/tools/gyp/v8.gyp#newcode63
tools/gyp/v8.gyp:63: 'xcode_settings': {'OTHER_LDFLAGS': ['-dynamiclib',
'-all_load']},
On 2012/03/28 16:08:51, Jakob wrote:
> 80 cols per line, please.
Ops, sorry :)

http://codereview.chromium.org/9808065/diff/5001/tools/gyp/v8.gyp#newcode542
tools/gyp/v8.gyp:542: ['v8_target_arch=="ia32"', {
On 2012/03/28 16:08:51, Jakob wrote:
> From crbug.com/44205, I gather that it should be possible to choose the
> architecture from within Xcode; v8_target_arch=="mac" (or target_arch=="mac",
I
> don't care) seems to be how the desire to have this option is indicated. Can
you
> make that work?
> Alternatively, would you mind leaving the 'v8_target_arch=="mac"' condition in
> so that any progress in that area is not hampered?

Ok, I'll leave this as it was before. I found it weird that both the ia32 and
x64 versions were being compiled and assembled into the same .a file (which
worked fine, I just felt odd and thought it could be an unintentional
condition).

[fdmanana, 2012-03-28 16:51:02]

Patch set 4 addresses your review Jakob.
I tested this with all the combinations of make/xcode, ia32/x64, gcc/clang,
release/debug mode. Seems to work well as before.

[fdmanana, 2012-03-30 14:16:35]

Rebased against latest bleeding_edge branch and re-tested.

[Jakob Kummerow, 2012-03-30 15:00:36]

Thanks. Landing this now.

I took the liberty to do a minor refactoring to common.gypi.

Sorry for the delay.

[Jakob Kummerow, 2012-03-30 15:00:55]

forgot: LGTM :-)

[fdmanana, 2012-03-30 15:52:03]

Thanks Jakob!