Fix gyptest-xcode-gcc.py when using Xcode 5.0.2

Fix gyptest-xcode-gcc.py when using Xcode 5.0.2

With Xcode 5.0.2, gcc is now a special version of clang and it no
longer supports the -Wno-invalid-offsetof. Skip the test then.

BUG=381
R=mark@chromium.org

Committed: https://code.google.com/p/gyp/source/detail?r=1809

[sdefresne, 2013-11-22 09:24:06]

Please take a look.

[Mark Mentovai, 2013-12-06 20:17:56]

https://codereview.chromium.org/83113002/diff/1/test/mac/gyptest-xcode-gcc.py
File test/mac/gyptest-xcode-gcc.py (right):

https://codereview.chromium.org/83113002/diff/1/test/mac/gyptest-xcode-gcc.py...
test/mac/gyptest-xcode-gcc.py:34: return test.format != 'xcode' or
os.readlink('/usr/bin/cc') != 'clang'
I think that testing |'clang' in CompilerVersion('/usr/bin/gcc')| is a better
version of the os.readlink test, because it should detect both symbolic link and
non-symbolic-link installations of cc (or gcc) as clang. Can you simplify this
function?

[sdefresne, 2013-12-09 16:11:11]

PTAL

[Mark Mentovai, 2013-12-09 16:29:05]

https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
File test/mac/gyptest-xcode-gcc.py (right):

https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
test/mac/gyptest-xcode-gcc.py:25: # "make" and "ninja" tests are hard-coded to
use "gcc/g++" but Xcode has some
This comment is stale. ninja uses cc and c++. Nobody cares what make uses as
long as this test passes. You can get rid of default_compiler and hard-code
'/usr/bin/cc'.

[sdefresne, 2013-12-10 12:38:43]

https://codereview.chromium.org/83113002/diff/1/test/mac/gyptest-xcode-gcc.py
File test/mac/gyptest-xcode-gcc.py (right):

https://codereview.chromium.org/83113002/diff/1/test/mac/gyptest-xcode-gcc.py...
test/mac/gyptest-xcode-gcc.py:34: return test.format != 'xcode' or
os.readlink('/usr/bin/cc') != 'clang'
On 2013/12/06 20:17:56, Mark Mentovai wrote:
> I think that testing |'clang' in CompilerVersion('/usr/bin/gcc')| is a better
> version of the os.readlink test, because it should detect both symbolic link
and
> non-symbolic-link installations of cc (or gcc) as clang. Can you simplify this
> function?

Done.

https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
File test/mac/gyptest-xcode-gcc.py (right):

https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
test/mac/gyptest-xcode-gcc.py:25: # "make" and "ninja" tests are hard-coded to
use "gcc/g++" but Xcode has some
On 2013/12/09 16:29:06, Mark Mentovai wrote:
> This comment is stale. ninja uses cc and c++. Nobody cares what make uses as
> long as this test passes. You can get rid of default_compiler and hard-code
> '/usr/bin/cc'.

Looking at test/mac/xcode-gcc/test.gyp, I still see them overriding the
compilers to be /usr/bin/{gcc,g++}:

  'make_global_settings': [
    ['CC', '/usr/bin/gcc'],
    ['CXX', '/usr/bin/g++'],
  ],

This settings appears to be used by the generated mkefile / ninja build files
but is not used by the Xcode generated project. So I think that on the bot, the
compiler used will still be /usr/bin/gcc that is a proper gcc.

Do you want me to also remove the override from test.gyp?

[Mark Mentovai, 2013-12-10 14:11:25]

https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
File test/mac/gyptest-xcode-gcc.py (right):

https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
test/mac/gyptest-xcode-gcc.py:25: # "make" and "ninja" tests are hard-coded to
use "gcc/g++" but Xcode has some
sdefresne wrote:
> On 2013/12/09 16:29:06, Mark Mentovai wrote:
> > This comment is stale. ninja uses cc and c++. Nobody cares what make uses as
> > long as this test passes. You can get rid of default_compiler and hard-code
> > '/usr/bin/cc'.
> 
> Looking at test/mac/xcode-gcc/test.gyp, I still see them overriding the
> compilers to be /usr/bin/{gcc,g++}:
> 
>   'make_global_settings': [
>     ['CC', '/usr/bin/gcc'],
>     ['CXX', '/usr/bin/g++'],
>   ],
> 
> This settings appears to be used by the generated mkefile / ninja build files
> but is not used by the Xcode generated project. So I think that on the bot,
the
> compiler used will still be /usr/bin/gcc that is a proper gcc.
> 
> Do you want me to also remove the override from test.gyp?

I think it’s safe to get rid of that now.

[sdefresne, 2013-12-10 16:06:18]

On 2013/12/10 14:11:25, Mark Mentovai wrote:
>
https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
> File test/mac/gyptest-xcode-gcc.py (right):
> 
>
https://codereview.chromium.org/83113002/diff/40001/test/mac/gyptest-xcode-gc...
> test/mac/gyptest-xcode-gcc.py:25: # "make" and "ninja" tests are hard-coded to
> use "gcc/g++" but Xcode has some
> sdefresne wrote:
> > On 2013/12/09 16:29:06, Mark Mentovai wrote:
> > > This comment is stale. ninja uses cc and c++. Nobody cares what make uses
as
> > > long as this test passes. You can get rid of default_compiler and
hard-code
> > > '/usr/bin/cc'.
> > 
> > Looking at test/mac/xcode-gcc/test.gyp, I still see them overriding the
> > compilers to be /usr/bin/{gcc,g++}:
> > 
> >   'make_global_settings': [
> >     ['CC', '/usr/bin/gcc'],
> >     ['CXX', '/usr/bin/g++'],
> >   ],
> > 
> > This settings appears to be used by the generated mkefile / ninja build
files
> > but is not used by the Xcode generated project. So I think that on the bot,
> the
> > compiler used will still be /usr/bin/gcc that is a proper gcc.
> > 
> > Do you want me to also remove the override from test.gyp?
> 
> I think it’s safe to get rid of that now.

Done, ptal.

[Mark Mentovai, 2013-12-10 16:32:28]

LGTM

[sdefresne, 2013-12-10 17:45:27]

Committed patchset #3 manually as r1809 (presubmit successful).