Image filters: fix srcOffset handling in asFragmentProcessor() path.

Image filters: fix srcOffset handling in asFragmentProcessor() path.

[Cherry-pick of eae84c2e0e2126374cd488a1c8a3e18169145635 to M49 branch.]

Filters such as SkMatrixConvolutionImageFilter which use the
asFragmentProcessor() path were not correctly handling srcOffset.
It is correctly applied to the bounds, but the srcRect and dstRect
were computed from the pre-offset bounds.

The fix is to move them to just above where they're used in drawing.

Note: this change adds a new test case to the imagefiltersgraph GM,
so it will have to be rebaselined post-landing.

BUG=skia:4855, 581389
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1637443003

NOTREECHECKS=true
NOTRY=true
NOPRESUBMIT=true
TBR=robertphillips@google.com

Committed: https://skia.googlesource.com/skia/+/109a6d5c08c078a110812628ea35aa2ca85c3489

[Stephen White, 2016-01-29 14:06:47]

Description was changed from

==========
Image filters: fix srcOffset handling in asFragmentProcessor() path.

[Cherry-pick of eae84c2e0e2126374cd488a1c8a3e18169145635 to M49 branch.]

Filters such as SkMatrixConvolutionImageFilter which use the
asFragmentProcessor() path were not correctly handling srcOffset.
It is correctly applied to the bounds, but the srcRect and dstRect
were computed from the pre-offset bounds.

The fix is to move them to just above where they're used in drawing.

Note: this change adds a new test case to the imagefiltersgraph GM,
so it will have to be rebaselined post-landing.

BUG=skia:4855
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Review URL: https://codereview.chromium.org/1637443003
==========

to

==========
Image filters: fix srcOffset handling in asFragmentProcessor() path.

[Cherry-pick of eae84c2e0e2126374cd488a1c8a3e18169145635 to M49 branch.]

Filters such as SkMatrixConvolutionImageFilter which use the
asFragmentProcessor() path were not correctly handling srcOffset.
It is correctly applied to the bounds, but the srcRect and dstRect
were computed from the pre-offset bounds.

The fix is to move them to just above where they're used in drawing.

Note: this change adds a new test case to the imagefiltersgraph GM,
so it will have to be rebaselined post-landing.

BUG=skia:4855
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Review URL: https://codereview.chromium.org/1637443003
NOTREECHECKS=true
NOTRY=true
NOPRESUBMIT=true
==========

[Stephen White, 2016-01-29 14:08:33]

Description was changed from

==========
Image filters: fix srcOffset handling in asFragmentProcessor() path.

[Cherry-pick of eae84c2e0e2126374cd488a1c8a3e18169145635 to M49 branch.]

Filters such as SkMatrixConvolutionImageFilter which use the
asFragmentProcessor() path were not correctly handling srcOffset.
It is correctly applied to the bounds, but the srcRect and dstRect
were computed from the pre-offset bounds.

The fix is to move them to just above where they're used in drawing.

Note: this change adds a new test case to the imagefiltersgraph GM,
so it will have to be rebaselined post-landing.

BUG=skia:4855
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

Review URL: https://codereview.chromium.org/1637443003
NOTREECHECKS=true
NOTRY=true
NOPRESUBMIT=true
==========

to

==========
Image filters: fix srcOffset handling in asFragmentProcessor() path.

[Cherry-pick of eae84c2e0e2126374cd488a1c8a3e18169145635 to M49 branch.]

Filters such as SkMatrixConvolutionImageFilter which use the
asFragmentProcessor() path were not correctly handling srcOffset.
It is correctly applied to the bounds, but the srcRect and dstRect
were computed from the pre-offset bounds.

The fix is to move them to just above where they're used in drawing.

Note: this change adds a new test case to the imagefiltersgraph GM,
so it will have to be rebaselined post-landing.

BUG=skia:4855,581389
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

NOTREECHECKS=true
NOTRY=true
NOPRESUBMIT=true
TBR=robertphillips@google.com
==========

[Stephen White, 2016-01-29 14:08:33]

senorblanco@chromium.org changed reviewers:
+ robertphillips@google.com

[Stephen White, 2016-01-29 14:08:52]

Description was changed from

==========
Image filters: fix srcOffset handling in asFragmentProcessor() path.

[Cherry-pick of eae84c2e0e2126374cd488a1c8a3e18169145635 to M49 branch.]

Filters such as SkMatrixConvolutionImageFilter which use the
asFragmentProcessor() path were not correctly handling srcOffset.
It is correctly applied to the bounds, but the srcRect and dstRect
were computed from the pre-offset bounds.

The fix is to move them to just above where they're used in drawing.

Note: this change adds a new test case to the imagefiltersgraph GM,
so it will have to be rebaselined post-landing.

BUG=skia:4855,581389
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

NOTREECHECKS=true
NOTRY=true
NOPRESUBMIT=true
TBR=robertphillips@google.com
==========

to

==========
Image filters: fix srcOffset handling in asFragmentProcessor() path.

[Cherry-pick of eae84c2e0e2126374cd488a1c8a3e18169145635 to M49 branch.]

Filters such as SkMatrixConvolutionImageFilter which use the
asFragmentProcessor() path were not correctly handling srcOffset.
It is correctly applied to the bounds, but the srcRect and dstRect
were computed from the pre-offset bounds.

The fix is to move them to just above where they're used in drawing.

Note: this change adds a new test case to the imagefiltersgraph GM,
so it will have to be rebaselined post-landing.

BUG=skia:4855,581389
GOLD_TRYBOT_URL=
https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&is...

NOTREECHECKS=true
NOTRY=true
NOPRESUBMIT=true
TBR=robertphillips@google.com

Committed:
https://skia.googlesource.com/skia/+/109a6d5c08c078a110812628ea35aa2ca85c3489
==========

[Stephen White, 2016-01-29 14:08:53]

Committed patchset #1 (id:1) manually as
109a6d5c08c078a110812628ea35aa2ca85c3489 (presubmit successful).

[robertphillips, 2016-01-29 17:34:23]

lgtm