Removed global static initializer redefines for android build.

Removed global static initializer redefines for android build.

[chudy, 2012-08-13 16:55:30]

Part 2 of 2 part CL to remove warnings from android build. Part 1 located at https://codereview.appspot.com/6446128 https://codereview.appspot.com/6448146/diff/1/include/core/SkPostConfig.h File include/core/SkPostConfig.h (left): https://codereview.appspot.com/6448146/diff/1/include/core/SkPostConfig.h#old... include/core/SkPostConfig.h:326: #ifndef SK_ALLOW_STATIC_GLOBAL_INITIALIZERS Redundant code since this is now defined by default to 1 in the common_conditions and common_variables gyps

[TomH, 2012-08-13 16:59:18]

On 2012/08/13 16:55:30, chudy wrote: > Part 2 of 2 part CL to remove warnings from android build. > > Part 1 located at https://codereview.appspot.com/6446128 > > https://codereview.appspot.com/6448146/diff/1/include/core/SkPostConfig.h > File include/core/SkPostConfig.h (left): > > https://codereview.appspot.com/6448146/diff/1/include/core/SkPostConfig.h#old... > include/core/SkPostConfig.h:326: #ifndef SK_ALLOW_STATIC_GLOBAL_INITIALIZERS > Redundant code since this is now defined by default to 1 in the > common_conditions and common_variables gyps I don't follow. Chrome #defines it to 0, because they don't allow any static globals for any of their code. We need to allow them to do so. Is your change compatible with their way of doing this, or do we need to change their code too?

[TomH, 2012-08-13 17:09:53]

It looks like Chrome sets SK_ALLOW_STATIC_GLOBAL_INITIALIZERS=0 in skia/skia.gyp. I thought that they were *supposed* to set it in skia/config/SkUserConfig.h. Have we changed the convention? Are they doing it wrong? Should we be supporting both approaches?

[TomH, 2012-08-13 17:10:15]

(change which Mike gets added)

[chudy, 2012-08-13 17:10:28]

As of right now this is compatible because chromium builds skia with their own stand-alone gyp file that defines SK_ALLOW_STATIC_GLOBAL_INITIALIZERS=0 on line 839. There is no collision On Mon, Aug 13, 2012 at 12:59 PM, <tomhudson@google.com> wrote: > > On 2012/08/13 16:55:30, chudy wrote: >> >> Part 2 of 2 part CL to remove warnings from android build. > > >> Part 1 located at https://codereview.appspot.com/6446128 > > > > https://codereview.appspot.com/6448146/diff/1/include/core/SkPostConfig.h >> >> File include/core/SkPostConfig.h (left): > > > > https://codereview.appspot.com/6448146/diff/1/include/core/SkPostConfig.h#old... >> >> include/core/SkPostConfig.h:326: #ifndef > > SK_ALLOW_STATIC_GLOBAL_INITIALIZERS >> >> Redundant code since this is now defined by default to 1 in the >> common_conditions and common_variables gyps > > > I don't follow. Chrome #defines it to 0, because they don't allow any > static globals for any of their code. We need to allow them to do so. Is > your change compatible with their way of doing this, or do we need to > change their code too? > > https://codereview.appspot.com/6448146/ -- Karol Chudy Phone: (203) 898-4233 chudy@google.com 

[bsalomon, 2012-08-13 17:23:44]

We shouldn't remove the SkPostConfig bit. Otherwise, the macro won't ever get defined when Skia is built using a build system other than gyp. It should be safe to leave it since it first checks whether the macro is already defined (as it will be in our GYP system).

[reed1, 2012-08-13 17:32:20]

My original design was that the client could freely use either SkUserConfig.h OR their own build-system defines, or any combination. All that Skia had to promise was: 1. preconfig would set up defaults 2. postconfig would ensure that we landed in an interpretable/consistent state.