Make it possible to generate the ios xcode proj by specifying only skia_os="ios" in GYP_DEFINES.

Make it possible to generate the ios xcode proj by specifying only skia_os="ios" in GYP_DEFINES.

R=caryclark@google.com,epoger@google.com
Committed: https://code.google.com/p/skia/source/detail?r=5796

[caryclark1, 2012-10-03 19:52:24]

lgtm

[bsalomon, 2012-10-03 19:54:55]

Adding Eric. Eric will this adversely affect Android? If the scripts that
process our gyp files for android have already specified armv7 and arm_neon then
I think we're OK since they are specified here with the % modifier.

[EricB, 2012-10-03 20:00:51]

This shouldn't break Android.  We set these variables in our build scripts.  We
should probably move that logic into gyp though, since it's a bit confusing to
have it in two places.

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi
File gyp/common_variables.gypi (right):

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi#newco...
gyp/common_variables.gypi:50: },
Do we really need this fourth level?

[bsalomon, 2012-10-03 20:04:37]

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi
File gyp/common_variables.gypi (right):

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi#newco...
gyp/common_variables.gypi:50: },
On 2012/10/03 20:00:51, EricB wrote:
> Do we really need this fourth level?

Yeah, I want skia_arch_type to depend on skia_os. Both were previously defined
at level 3. So I had to make skia_os be defined at level 4.  Isn't gyp lovely?

[EricB, 2012-10-03 20:08:59]

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi
File gyp/common_variables.gypi (right):

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi#newco...
gyp/common_variables.gypi:50: },
On 2012/10/03 20:04:37, bsalomon wrote:
> On 2012/10/03 20:00:51, EricB wrote:
> > Do we really need this fourth level?
> 
> Yeah, I want skia_arch_type to depend on skia_os. Both were previously defined
> at level 3. So I had to make skia_os be defined at level 4.  Isn't gyp lovely?

But do we use it at level 3?  It seems like it could be at level 2, right where
skia_arch_width has its os-dependent switch.

[bsalomon, 2012-10-03 20:16:01]

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi
File gyp/common_variables.gypi (right):

https://codereview.appspot.com/6601051/diff/1/gyp/common_variables.gypi#newco...
gyp/common_variables.gypi:50: },
On 2012/10/03 20:08:59, EricB wrote:
> On 2012/10/03 20:04:37, bsalomon wrote:
> > On 2012/10/03 20:00:51, EricB wrote:
> > > Do we really need this fourth level?
> > 
> > Yeah, I want skia_arch_type to depend on skia_os. Both were previously
defined
> > at level 3. So I had to make skia_os be defined at level 4.  Isn't gyp
lovely?
> 
> But do we use it at level 3?  It seems like it could be at level 2, right
where
> skia_arch_width has its os-dependent switch.

Ah-ha. I think you're right. It appears that in the current file skia_arch_type
is unnecessarily defined in the 3rd circle of variable hell and could have been
defined at 2 instead or even 1 since no other var depends on it.

I'll post a revised change.

[bsalomon, 2012-10-03 20:28:27]

The latest patch moves skia_arch_type to level 1.

https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi
File gyp/common_variables.gypi (right):

https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi#ne...
gyp/common_variables.gypi:106: 'armv7%': 1,
Because armv7 and arm_neon are now at level 1 and don't need to be rippled up
the variables dicts it isn't required for them to defined in the else case.

[EricB, 2012-10-03 20:33:22]

https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi
File gyp/common_variables.gypi (right):

https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi#ne...
gyp/common_variables.gypi:106: 'armv7%': 1,
On 2012/10/03 20:28:28, bsalomon wrote:
> Because armv7 and arm_neon are now at level 1 and don't need to be rippled up
> the variables dicts it isn't required for them to defined in the else case.

Sorry to nag, but are we sure that this compiles?  It looks like skia_arch_type
is used in level 2 dicts in common_conditions.gypi, although maybe I'm
misunderstanding what the levels mean.

[bsalomon, 2012-10-03 20:56:08]

On 2012/10/03 20:33:22, EricB wrote:
> https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi
> File gyp/common_variables.gypi (right):
> 
>
https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi#ne...
> gyp/common_variables.gypi:106: 'armv7%': 1,
> On 2012/10/03 20:28:28, bsalomon wrote:
> > Because armv7 and arm_neon are now at level 1 and don't need to be rippled
up
> > the variables dicts it isn't required for them to defined in the else case.
> 
> Sorry to nag, but are we sure that this compiles?  It looks like
skia_arch_type
> is used in level 2 dicts in common_conditions.gypi, although maybe I'm
> misunderstanding what the levels mean.

No prob... this stuff is confusing as all hell. I believe that  it is OK because
the usage in common_conditions is within a conditions dictionary not a variables
dictionary. I was able to build Mac and iOS SampleApp with the change.

[EricB, 2012-10-03 20:57:18]

On 2012/10/03 20:56:08, bsalomon wrote:
> On 2012/10/03 20:33:22, EricB wrote:
> > https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi
> > File gyp/common_variables.gypi (right):
> > 
> >
>
https://codereview.appspot.com/6601051/diff/5003/gyp/common_variables.gypi#ne...
> > gyp/common_variables.gypi:106: 'armv7%': 1,
> > On 2012/10/03 20:28:28, bsalomon wrote:
> > > Because armv7 and arm_neon are now at level 1 and don't need to be rippled
> up
> > > the variables dicts it isn't required for them to defined in the else
case.
> > 
> > Sorry to nag, but are we sure that this compiles?  It looks like
> skia_arch_type
> > is used in level 2 dicts in common_conditions.gypi, although maybe I'm
> > misunderstanding what the levels mean.
> 
> No prob... this stuff is confusing as all hell. I believe that  it is OK
because
> the usage in common_conditions is within a conditions dictionary not a
variables
> dictionary. I was able to build Mac and iOS SampleApp with the change.

LGTM