Split radial_gradients2 GM into two to reproduce its bug.

Split radial_gradients2 GM into two to reproduce its bug.

Android HW rendering seems to be taking the premul-first path, while
Skia takes the interpolate-first path (both SW and HW).

BUG=
R=reed@google.com

Committed: https://code.google.com/p/skia/source/detail?r=11284

[reed1, 2013-09-11 14:33:27]

https://codereview.chromium.org/23637008/diff/1/gm/gradients.cpp
File gm/gradients.cpp (right):

https://codereview.chromium.org/23637008/diff/1/gm/gradients.cpp#newcode446
gm/gradients.cpp:446: static GM* MyFactory7(void*) { return new
RadialGradient2GM(true); }
nitties:

1. If we're going to have 2 gms, might be slightly clearer if we just pass the
"flags" for the gradient as a param, rather than a bool which we happen to know
translates into diff flags.

2. Can we instead change the curr GM to just draw twice on screen, so we can see
at a glance the two different effects side-by-side?

questions, not demands

[mtklein, 2013-09-11 14:50:56]

On 2013/09/11 14:33:27, reed1 wrote:
> https://codereview.chromium.org/23637008/diff/1/gm/gradients.cpp
> File gm/gradients.cpp (right):
> 
> https://codereview.chromium.org/23637008/diff/1/gm/gradients.cpp#newcode446
> gm/gradients.cpp:446: static GM* MyFactory7(void*) { return new
> RadialGradient2GM(true); }
> nitties:
> 
> 1. If we're going to have 2 gms, might be slightly clearer if we just pass the
> "flags" for the gradient as a param, rather than a bool which we happen to
know
> translates into diff flags.
> 
> 2. Can we instead change the curr GM to just draw twice on screen, so we can
see
> at a glance the two different effects side-by-side?
> 
> questions, not demands

Yep, 2 sounds best.  Will do.

[mtklein, 2013-09-11 14:56:13]

On 2013/09/11 14:50:56, mtklein wrote:
> On 2013/09/11 14:33:27, reed1 wrote:
> > https://codereview.chromium.org/23637008/diff/1/gm/gradients.cpp
> > File gm/gradients.cpp (right):
> > 
> > https://codereview.chromium.org/23637008/diff/1/gm/gradients.cpp#newcode446
> > gm/gradients.cpp:446: static GM* MyFactory7(void*) { return new
> > RadialGradient2GM(true); }
> > nitties:
> > 
> > 1. If we're going to have 2 gms, might be slightly clearer if we just pass
the
> > "flags" for the gradient as a param, rather than a bool which we happen to
> know
> > translates into diff flags.
> > 
> > 2. Can we instead change the curr GM to just draw twice on screen, so we can
> see
> > at a glance the two different effects side-by-side?
> > 
> > questions, not demands
> 
> Yep, 2 sounds best.  Will do.

ptal

[reed1, 2013-09-11 15:03:11]

lgtm

[mtklein, 2013-09-16 13:02:57]

Committed patchset #2 manually as r11284 (presubmit successful).