[PDF] Fix shaders when the initial transform has a translation.

[PDF] Fix shaders when the initial transform has a translation.

Fixes chromium bug 115041

Committed: https://code.google.com/p/skia/source/detail?r=3420

[Steve VanDeBogart, 2012-03-16 17:39:10]

https://codereview.appspot.com/5849045/diff/1/gm/shaderbounds.cpp
File gm/shaderbounds.cpp (right):

https://codereview.appspot.com/5849045/diff/1/gm/shaderbounds.cpp#newcode30
gm/shaderbounds.cpp:30: ShaderBoundsGM(ShaderGenFunc maker, const SkString&
name)
I made this gm more general than needed to fix this problem because there are
other related problems to be fixed, but I don't have time to work on them at the
moment.

https://codereview.appspot.com/5849045/diff/1/gyp/gmslides.gypi
File gyp/gmslides.gypi (right):

https://codereview.appspot.com/5849045/diff/1/gyp/gmslides.gypi#newcode40
gyp/gmslides.gypi:40: '../gm/shaderbounds.cpp',
There's a process to get new baselines now, right?  Can you point me toward it?

https://codereview.appspot.com/5849045/diff/1/src/pdf/SkPDFDevice.cpp
File src/pdf/SkPDFDevice.cpp (right):

https://codereview.appspot.com/5849045/diff/1/src/pdf/SkPDFDevice.cpp#newcode...
src/pdf/SkPDFDevice.cpp:1412: // We need to apply the initial transform to
bounds in order to get
This is the actual fix.

[reed1, 2012-03-16 18:24:04]

Steve, I'm not sure why we need to add something to gm. Is this required to make
the raster and gpu versions draw the same as the PDF version?

[Steve VanDeBogart, 2012-03-16 18:27:01]

On 2012/03/16 18:24:04, reed1 wrote:
> Steve, I'm not sure why we need to add something to gm. Is this required to
make
> the raster and gpu versions draw the same as the PDF version?

The problem only happens when the pdf initial transform has a translation.  The
current gm sets the the initial transform to identity.  The modifications to gm
add the infrastructure so that the gm can specify the initial transform.

[reed1, 2012-03-16 18:47:38]

Does this CL fix any pre-existing gm, or is it only adding a new testing mode
for testing new gms?

[Steve VanDeBogart, 2012-03-16 18:49:56]

On 2012/03/16 18:47:38, reed1 wrote:
> Does this CL fix any pre-existing gm, or is it only adding a new testing mode
> for testing new gms?

No existing gm tests the problem fixed in SkPDFDevice.cpp in this CL.  I added a
new gm to test that problem as well as the infrastructure needed for that gm. 
That gm doesn't affect any existing gm tests, only the new test.  I can break
the change into the different parts if you like.

[Steve VanDeBogart, 2012-03-16 18:52:21]

On 2012/03/16 18:49:56, Steve VanDeBogart wrote:
> On 2012/03/16 18:47:38, reed1 wrote:
> > Does this CL fix any pre-existing gm, or is it only adding a new testing
mode
> > for testing new gms?
> 
> No existing gm tests the problem fixed in SkPDFDevice.cpp in this CL.  I added
a
> new gm to test that problem as well as the infrastructure needed for that gm. 
> That gm doesn't affect any existing gm tests, only the new test.  I can break
> the change into the different parts if you like.

Sorry - that gm infrastructure change doesn't affect any existing gm tests.

[reed1, 2012-03-16 19:04:24]

Lets check in just SkPDFDevice.cpp for now (lgtm)

Next week, I'd like to have a chat w/ you, me, and ben, to talk about:

1. What is the best way to test additional APIs that PDF has (e.g.
initialTransform)
2. What is the best API for Skia to have to make printing (1) great, (2)
clear/consistent

[Steve VanDeBogart, 2012-03-16 19:10:37]

On 2012/03/16 19:04:24, reed1 wrote:
> Lets check in just SkPDFDevice.cpp for now (lgtm)
> 
> Next week, I'd like to have a chat w/ you, me, and ben, to talk about:
> 
> 1. What is the best way to test additional APIs that PDF has (e.g.
> initialTransform)
> 2. What is the best API for Skia to have to make printing (1) great, (2)
> clear/consistent

I scheduled a meeting for Tuesday, reschedule if you like.