Add support for new glShaderSource() signature from Khronos. Disabled by default.

Add support for new glShaderSource() signature from Khronos.  Disabled by default.

Committed: https://code.google.com/p/skia/source/detail?r=6491

[Stephen White, 2012-11-19 16:49:42]

PTAL.  Apparently this is required to update to the latest headers from Khronos.

[bsalomon, 2012-11-19 17:17:29]

On 2012/11/19 16:49:42, Stephen White wrote:
> PTAL.  Apparently this is required to update to the latest headers from
Khronos.

Thanks for fixing this. I didn't know they had changed the sig. Can you change
the macro to GR_<blah> rather than SK_?  Also, in the GR code we use macros that
are always defined (and #if check their values rather than #ifdef). The usual
pattern is to add:

#if !defined(GR_THING)
   #define GR_THING  <default value, usually 0 or 1>
#endif

in GrConfig.h in the post config section. It keeps all the possible GR_ macros
listed/documented in one file.

[Stephen White, 2012-11-19 18:19:27]

On 2012/11/19 17:17:29, bsalomon wrote:
> On 2012/11/19 16:49:42, Stephen White wrote:
> > PTAL.  Apparently this is required to update to the latest headers from
> Khronos.
> 
> Thanks for fixing this. I didn't know they had changed the sig. Can you change
> the macro to GR_<blah> rather than SK_?  Also, in the GR code we use macros
that
> are always defined (and #if check their values rather than #ifdef). The usual
> pattern is to add:
> 
> #if !defined(GR_THING)
>    #define GR_THING  <default value, usually 0 or 1>
> #endif
> 
> in GrConfig.h in the post config section. It keeps all the possible GR_ macros
> listed/documented in one file.

Sure.  I was actually hoping we could remove this #define once Chrome had rolled
and switched headers, but I'll clean it up just in case we can't.

[bsalomon, 2012-11-19 18:26:27]

On 2012/11/19 18:19:27, Stephen White wrote:
> On 2012/11/19 17:17:29, bsalomon wrote:
> > On 2012/11/19 16:49:42, Stephen White wrote:
> > > PTAL.  Apparently this is required to update to the latest headers from
> > Khronos.
> > 
> > Thanks for fixing this. I didn't know they had changed the sig. Can you
change
> > the macro to GR_<blah> rather than SK_?  Also, in the GR code we use macros
> that
> > are always defined (and #if check their values rather than #ifdef). The
usual
> > pattern is to add:
> > 
> > #if !defined(GR_THING)
> >    #define GR_THING  <default value, usually 0 or 1>
> > #endif
> > 
> > in GrConfig.h in the post config section. It keeps all the possible GR_
macros
> > listed/documented in one file.
> 
> Sure.  I was actually hoping we could remove this #define once Chrome had
rolled
> and switched headers, but I'll clean it up just in case we can't.

Oh got it.. but wont we have to keep the old sig for native GrGLInterfaces
(until platforms update their headers)?

[Stephen White, 2012-11-19 18:34:40]

On 2012/11/19 18:26:27, bsalomon wrote:
> On 2012/11/19 18:19:27, Stephen White wrote:
> > On 2012/11/19 17:17:29, bsalomon wrote:
> > > On 2012/11/19 16:49:42, Stephen White wrote:
> > > > PTAL.  Apparently this is required to update to the latest headers from
> > > Khronos.
> > > 
> > > Thanks for fixing this. I didn't know they had changed the sig. Can you
> change
> > > the macro to GR_<blah> rather than SK_?  Also, in the GR code we use
macros
> > that
> > > are always defined (and #if check their values rather than #ifdef). The
> usual
> > > pattern is to add:
> > > 
> > > #if !defined(GR_THING)
> > >    #define GR_THING  <default value, usually 0 or 1>
> > > #endif
> > > 
> > > in GrConfig.h in the post config section. It keeps all the possible GR_
> macros
> > > listed/documented in one file.
> > 
> > Sure.  I was actually hoping we could remove this #define once Chrome had
> rolled
> > and switched headers, but I'll clean it up just in case we can't.
> 
> Oh got it.. but wont we have to keep the old sig for native GrGLInterfaces
> (until platforms update their headers)?

Since Skia has an implementation that puts desktop GL's glShaderSource into a
GrGLInterface slot, I think we'll just have to cast that once we switch.  It's
just a "const" on something that was presumably already being treated as const,
so it should be safe.. I hope.

[Stephen White, 2012-11-19 20:27:35]

Fixed re: review comments.  PTAL

[bsalomon, 2012-11-19 20:38:16]

On 2012/11/19 20:27:35, Stephen White wrote:
> Fixed re: review comments.  PTAL

LGTM.