platform/upstream/libSkiaSharp.git
8 years agoAdd gm for displaying different variants of 8888 bitmaps and fix a small rejection...
herb [Fri, 3 Jun 2016 16:44:53 +0000 (09:44 -0700)]
Add gm for displaying different variants of 8888 bitmaps and fix a small rejection case for sprite blitters.

BUG=skia:5378
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2036493004

Review-Url: https://codereview.chromium.org/2036493004

8 years agofix windows assert
caryclark [Fri, 3 Jun 2016 16:42:53 +0000 (09:42 -0700)]
fix windows assert

Windows asserts trigger out of memory stack traces that
are confusing. This permits windows to break more
directly, stopping the IDE at the assert and providing
a less confusing stacktrace outside of the IDE.

R=reed@google.com,herb@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2039443002

Review-Url: https://codereview.chromium.org/2039443002

8 years agoFix undefined behavior in libpng
scroggo [Fri, 3 Jun 2016 16:36:53 +0000 (09:36 -0700)]
Fix undefined behavior in libpng

Check for a null source before calling memcpy.

BUG=skia:5390
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2040433002

Review-Url: https://codereview.chromium.org/2040433002

8 years agoSkPDF: handle unpremul right
halcanary [Fri, 3 Jun 2016 15:57:03 +0000 (08:57 -0700)]
SkPDF: handle unpremul right

.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2037083002

Review-Url: https://codereview.chromium.org/2037083002

8 years agoAbandon offset support with texel buffers
csmartdalton [Fri, 3 Jun 2016 15:50:54 +0000 (08:50 -0700)]
Abandon offset support with texel buffers

We don't seem to require nonzero offsets for texel buffers at this
point in time, and requiring this feature greatly reduces the number
of desktop clients that can use texel buffers. If we find a use for
offsets later we can always add it back as a separate feature.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2036953002

Review-Url: https://codereview.chromium.org/2036953002

8 years agoAdd Softkey UIState to Viewer
liyuqian [Fri, 3 Jun 2016 15:47:23 +0000 (08:47 -0700)]
Add Softkey UIState to Viewer

We can use this to simulate any key/command on Android UI.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2035923002

Review-Url: https://codereview.chromium.org/2035923002

8 years agoSupport ARB extensions for indirect drawing
csmartdalton [Fri, 3 Jun 2016 15:28:47 +0000 (08:28 -0700)]
Support ARB extensions for indirect drawing

Adds support for the ARB extensions and fixes incorrect logic that was
being used on the GL version for indirect drawing.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2038743002

Review-Url: https://codereview.chromium.org/2038743002

8 years agoAdd color correction benchmark - with comparison to qcms
msarett [Fri, 3 Jun 2016 15:25:21 +0000 (08:25 -0700)]
Add color correction benchmark - with comparison to qcms

--colorImages is empty by default so this won't affect the runtime of
the bots.  To run locally, use --colorImages <path> and
--benchType skcolorcodec.

Two takeaways so far:
Color correction is (currently) slower than jpeg decodes.
Our reference solution is slower than qcms.

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2035793002

Review-Url: https://codereview.chromium.org/2035793002

8 years agoFix uses of SkImageInfo in BRD to preserve color space info
msarett [Fri, 3 Jun 2016 15:23:40 +0000 (08:23 -0700)]
Fix uses of SkImageInfo in BRD to preserve color space info

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2035483005

Review-Url: https://codereview.chromium.org/2035483005

8 years agoskiaserve: Add a flag to control hosted mode, i.e. when running on debugger.skia...
jcgregorio [Fri, 3 Jun 2016 14:26:22 +0000 (07:26 -0700)]
skiaserve: Add a flag to control hosted mode, i.e. when running on debugger.skia.org.

BUG=skia:
TBR=ethannicholas@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2036173002

Review-Url: https://codereview.chromium.org/2036173002

8 years agoAdd Android device type for NVIDIA Shield
borenet [Fri, 3 Jun 2016 14:01:32 +0000 (07:01 -0700)]
Add Android device type for NVIDIA Shield

BUG=skia:5185
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2036183003

Review-Url: https://codereview.chromium.org/2036183003

8 years agoMake SkPngCodec decode progressively.
scroggo [Thu, 2 Jun 2016 21:16:43 +0000 (14:16 -0700)]
Make SkPngCodec decode progressively.

This is a step towards using SkCodec in Chromium, where progressive
decoding is necessary.

Switch from using png_read_row (which expects all the data to be
available) to png_process_data, which uses callbacks when rows are
available.

Create a new API for SkCodec, which supports progressive decoding and
scanline decoding. Future changes will switch the other clients off of
startScanlineDecode and get/skip-Scanlines to the new API.

Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced
PNG images. In the new API, interlaced PNG fits kTopDown. Also remove
updateCurrScanline(), which was only used by the old implementation for
interlaced PNG.

DMSrcSink:
- In CodecSrc::kScanline_Mode, use the new method for scanline decoding
for the supported formats (just PNG and PNG-in-ICO for now).

fuzz.cpp:
- Remove reference to kNone_ScanlineOrder

SkCodec:
- Add new APIs:
    - startIncrementalDecode
    - incrementalDecode
- Remove kNone_SkScanlineOrder and updateCurrScanline()

SkPngCodec:
- Implement new APIs
- Switch from sk_read_fn/png_read_row etc to png_process_data
- Expand AutoCleanPng's role to decode the header and create the
  SkPngCodec
- Make the interlaced PNG decoder report how many lines were
  initialized during an incomplete decode
- Make initializeSwizzler return a bool instead of an SkCodec::Result
  (It only returned kSuccess or kInvalidInput anyway)

SkIcoCodec:
- Implement the new APIs; supported for PNG in ICO

SkSampledCodec:
- Call the new method for decoding scanlines, and fall back to the old
  method if the new version is unimplemented
- Remove references to kNone_SkScanlineOrder

tests/CodecPartial:
- Add a test which decodes part of an image, then finishes the decode,
  and compares it to the straightforward method

tests/CodecTest:
- Add a test which decodes all scanlines using the new method
- Repurpose the Codec_stripes test to decode using the new method in
  sections rather than all at once
- In the method check(), add a parameter for whether the image supports
  the new method of scanline decoding, and be explicit about whether an
  image supports incomplete
- Test incomplete PNG decodes. We should have been doing it anyway for
  non-interlaced (except for an image that is too small - one row), but
  the new method supports interlaced incomplete as well
- Make test_invalid_parameters test the new method
- Add a test to ensure that it's safe to fall back to scanline decoding without
  rewinding

BUG=skia:4211

The new version was generally faster than the old version (but not significantly so).

Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/

Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003

Committed: https://skia.googlesource.com/skia/+/a4b09a117d4d1ba5dda372e6a2323e653766539e

Committed: https://skia.googlesource.com/skia/+/30e78c9737ff4861dc4e3fa1e4cd010680ed6965

Review-Url: https://codereview.chromium.org/1997703003

8 years agoCheck libpng directly into third_party/
scroggo [Thu, 2 Jun 2016 19:59:59 +0000 (12:59 -0700)]
Check libpng directly into third_party/

With this change, the CMake build, which does not use DEPS to sync
external projects, is able to build and use the same version of libpng
that is used in other builds.

This will allow all platforms (including Google3 CMake build) to test on
the same version of libpng, so we do not need to make SkPngCodec support
all versions of libpng.

- Update CMakeLists.txt to use the checked in libpng.
- Check in libpng version 1.6.22rc01
- Update README.google
- Replace our old LICENSE file with the latest one from libpng
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2033063003
CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Ubuntu-GCC-x86_64-Release-CMake-Trybot,Build-Mac-Clang-x86_64-Release-CMake-Trybot

Review-Url: https://codereview.chromium.org/2033063003

8 years agoSkMultiPictureDocument & SkMultiPictureDocumentReader
halcanary [Thu, 2 Jun 2016 19:41:14 +0000 (12:41 -0700)]
SkMultiPictureDocument & SkMultiPictureDocumentReader

also, a new DM::Src.

motivation: To be used to test the printing pipeline in Chromium.

BUG=skia:5370

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2023593002

Review-Url: https://codereview.chromium.org/2023593002

8 years agoadd MakeS32 helper to SkImageInfo, fix named-gamma constructor bug
reed [Thu, 2 Jun 2016 19:40:22 +0000 (12:40 -0700)]
add MakeS32 helper to SkImageInfo, fix named-gamma constructor bug

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2035813003

Review-Url: https://codereview.chromium.org/2035813003

8 years agoImplement OpenGL backend in Android viewer app
liyuqian [Thu, 2 Jun 2016 19:16:25 +0000 (12:16 -0700)]
Implement OpenGL backend in Android viewer app

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2032623002

Review-Url: https://codereview.chromium.org/2032623002

8 years agoFix argument parsing in android app launcher script
brianosman [Thu, 2 Jun 2016 18:44:06 +0000 (11:44 -0700)]
Fix argument parsing in android app launcher script

Makes the argument match the error message that gets printed (and users'
expectations).

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2030123002

Review-Url: https://codereview.chromium.org/2030123002

8 years agoTessellator: stop copying vertices into Polys and Monotones.
senorblanco [Thu, 2 Jun 2016 18:36:48 +0000 (11:36 -0700)]
Tessellator: stop copying vertices into Polys and Monotones.

The vertices which are produced by stage 5 of the
tesselator are copied into the Polys and MonotonePolys it
produces. This is necessary because each vertex may have an
arbitrary valence, since it may participate in an arbitrary
number of Polys, so we can't use the vertex's prev/next
pointers to represent all the Monotones of which this
vertex may be a member.

However, each Edge can only be a member of two Polys (one
on each side of the edge). So by adding two prev/next
pointer pairs to each Edge, we can represent each Monotone
as a list of edges instead. Then we no longer need to copy
the vertices.

One wrinkle is that the ear-clipping stage (6) of the
tessellator does require prev/next pointers, in order to
remove vertices as their ears are clipped. So we convert
the edge list into a vertex list during Monotone::emit(),
using the prev/next pointers temporarily for that monotone.

This change improves performance by 7-20% on a non-caching
version of the tessellator, and reduces memory use.

Other notes:

1) Polys are initially constructed empty (no edges), but
with the top vertex, which is needed for splitting
Polys. Edges are added to Polys only after their bottom
vertex is seen.

2) MonotonePolys are always constructed with one edge, so
we always know their handedness (left/right).
MonotonePoly::addEdge() no longer detects when a monotone
is "done" (edge of opposite handedness); this is handled
by Poly::addEdge(), so MonotonePoly::addEdge() has no
return value.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2029243002

Review-Url: https://codereview.chromium.org/2029243002

8 years agoAdjust tolerance for when Ganesh stroked-line-as-rect optimization kicks in
robertphillips [Thu, 2 Jun 2016 17:57:59 +0000 (10:57 -0700)]
Adjust tolerance for when Ganesh stroked-line-as-rect optimization kicks in

Some GPUs results look worse for the < 1.0f tolerance (i.e., iPad)

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2035843002

Review-Url: https://codereview.chromium.org/2035843002

8 years agoRemove SK_SUPPORT_LEGACY_TYPEFACE_MAKE_FROM_NAME.
bungeman [Thu, 2 Jun 2016 17:47:19 +0000 (10:47 -0700)]
Remove SK_SUPPORT_LEGACY_TYPEFACE_MAKE_FROM_NAME.

The define SK_SUPPORT_LEGACY_TYPEFACE_MAKE_FROM_NAME is no longer used
by any clients and the code it guards can now be removed as well.

TBR=reed
This just removes dead code.

Review-Url: https://codereview.chromium.org/2030983002

8 years agoExclude Mali devices from manual mip-mapping.
brianosman [Thu, 2 Jun 2016 17:19:10 +0000 (10:19 -0700)]
Exclude Mali devices from manual mip-mapping.

With manual mip-mapping, the Nexus 10 is failing the sRGB mip-map test. That suggests a failure in the sRGB implementation (decode after bilerp, possibly?). Using the driver's mip-mapper resolves the unit test issue, at least.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2036763002

Review-Url: https://codereview.chromium.org/2036763002

8 years agoAdd getting the size of a given mipmap level.
cblume [Thu, 2 Jun 2016 16:01:48 +0000 (09:01 -0700)]
Add getting the size of a given mipmap level.

When allocating memory for mipmaps, it would be very handy to know the
dimensions of a given mipmap level.

R=bsalomon@google.com
BUG=578304
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2018283002

Review-Url: https://codereview.chromium.org/2018283002

8 years agoAdd a caps bit to enable/disable the new manual mip-mapper
brianosman [Thu, 2 Jun 2016 15:59:34 +0000 (08:59 -0700)]
Add a caps bit to enable/disable the new manual mip-mapper

Turn it off on Adreno3xx devices, because the small-render-target bug on
those devices leads to empty mipmaps for the smallest mips.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2026393004

Review-Url: https://codereview.chromium.org/2026393004

8 years agoFix incorrect parameter in new benchmark.
brianosman [Thu, 2 Jun 2016 13:50:40 +0000 (06:50 -0700)]
Fix incorrect parameter in new benchmark.

An old value for one of the dimensions slipped in with the original CL.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2033713003

Review-Url: https://codereview.chromium.org/2033713003

8 years agoManually generated sRGB mipmaps, with successively smaller draws.
brianosman [Thu, 2 Jun 2016 12:49:21 +0000 (05:49 -0700)]
Manually generated sRGB mipmaps, with successively smaller draws.

Dirty GL-generated mipmaps whenever an sRGB texture is used with a new
value for TEXTURE_SRGB_DECODE. Add a new test rectangle to the gamma GM
that tests that textures are correctly converted to linear before
filtering when generating mipmaps.

Added a new unit test that alternates how a texture is interpreted (sRGB
or not), to verify that we rebuild mipmaps when needed, and that we get
the correct results out in both modes.

This test originally failed on four of our bots producing incorrect mips
in three different ways. I'm not real surprised, but it looks like
we can't rely on glGenerateMipmap to do the right thing, in conjunction
with TEXTURE_SRGB_DECODE.

Instead, actually create mip-chains using a series of draw calls.
(My first attempt used glBlitFramebuffer, and that still had bugs on
several bots). This approach appears to work correctly on any device
that fully supports sRGB.

Because the mipmap draws are fairly destructive to state, I had to
hoist them out of bindTexture. That means adding a second pass over
the texture accesses in the processor, at the very beginning of flush.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1840473002

Review-Url: https://codereview.chromium.org/2007973002

8 years agoHandle stroked single line special case in Ganesh
robertphillips [Thu, 2 Jun 2016 12:21:34 +0000 (05:21 -0700)]
Handle stroked single line special case in Ganesh

This CL roughly halves the time spent on the captured stroked lines skp.
On my Linux desktop it boosts the external benchmark from 2618 to 5007.

This is a companion to: https://codereview.chromium.org/2019193002/ (Add new GM to exercise stroked line special case)

The idea is to land the GM first so any regressions are visible.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2023693002

Review-Url: https://codereview.chromium.org/2023693002

8 years agoAdd a test to ensure that a case where SkCubicClipper::ChopMonoAtY returns false...
mbarbella [Wed, 1 Jun 2016 22:39:47 +0000 (15:39 -0700)]
Add a test to ensure that a case where SkCubicClipper::ChopMonoAtY returns false is handled properly.

Also fixes a style issue in the fix for the issue being tested.

BUG=chromium:613918
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2021343004

Review-Url: https://codereview.chromium.org/2021343004

8 years agoCreate SkColorSpaceXform to handle color conversions
msarett [Wed, 1 Jun 2016 21:47:18 +0000 (14:47 -0700)]
Create SkColorSpaceXform to handle color conversions

Also adds testing of qcms color correction, so we can compare
SkColorSpaceXform outputs to qcms outputs.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1952063002

Committed: https://skia.googlesource.com/skia/+/740cc88ee3d63c75e52d31238f2a32600cc57a8c

Review-Url: https://codereview.chromium.org/1952063002

8 years agoRecognize common parametric gamma
msarett [Wed, 1 Jun 2016 21:43:36 +0000 (14:43 -0700)]
Recognize common parametric gamma

5% of images with ICC profiles (in my test set of 10k)
have this exact gamma representation, which is very close
to 2.2f.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2027413002

Review-Url: https://codereview.chromium.org/2027413002

8 years agoRelease surface between runs to avoid seg fault
brianosman [Wed, 1 Jun 2016 20:49:27 +0000 (13:49 -0700)]
Release surface between runs to avoid seg fault

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2032693002

Review-Url: https://codereview.chromium.org/2032693002

8 years agoFix uninit warning on valgrind bot in SkPathPriv:IsSimpleClosedRect
Brian Salomon [Wed, 1 Jun 2016 20:02:17 +0000 (16:02 -0400)]
Fix uninit warning on valgrind bot in SkPathPriv:IsSimpleClosedRect

TBR=robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2030713002

Review URL: https://codereview.chromium.org/2030713002 .

8 years agoNew benchmark to measure GPU mip-mapping performance
brianosman [Wed, 1 Jun 2016 19:40:15 +0000 (12:40 -0700)]
New benchmark to measure GPU mip-mapping performance

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2030673002

Review-Url: https://codereview.chromium.org/2030673002

8 years agoRevert of Make SkPngCodec decode progressively. (patchset #24 id:480001 of https...
scroggo [Wed, 1 Jun 2016 19:08:23 +0000 (12:08 -0700)]
Revert of Make SkPngCodec decode progressively. (patchset #24 id:480001 of https://codereview.chromium.org/1997703003/ )

Reason for revert:
Still breaking Google3 e.g.:

https://sponge.corp.google.com/target?id=9261d31b-34fc-4f0f-981e-f92f7c5cea2c&target=//third_party/skia/HEAD:dm#shard=1|run=1|attempt=1|page=-1

https://test.corp.google.com/ui#cl=123773095&flags=CAMQBQ==&id=OCL:123773095:BASE:123773415:1464804876959:b0ea9b1c&t=//third_party/skia/HEAD:dm

Original issue's description:
> Make SkPngCodec decode progressively.
>
> This is a step towards using SkCodec in Chromium, where progressive
> decoding is necessary.
>
> Switch from using png_read_row (which expects all the data to be
> available) to png_process_data, which uses callbacks when rows are
> available.
>
> Create a new API for SkCodec, which supports progressive decoding and
> scanline decoding. Future changes will switch the other clients off of
> startScanlineDecode and get/skip-Scanlines to the new API.
>
> Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced
> PNG images. In the new API, interlaced PNG fits kTopDown. Also remove
> updateCurrScanline(), which was only used by the old implementation for
> interlaced PNG.
>
> DMSrcSink:
> - In CodecSrc::kScanline_Mode, use the new method for scanline decoding
> for the supported formats (just PNG and PNG-in-ICO for now).
>
> fuzz.cpp:
> - Remove reference to kNone_ScanlineOrder
>
> SkCodec:
> - Add new APIs:
>     - startIncrementalDecode
>     - incrementalDecode
> - Remove kNone_SkScanlineOrder and updateCurrScanline()
>
> SkPngCodec:
> - Implement new APIs
> - Switch from sk_read_fn/png_read_row etc to png_process_data
> - Expand AutoCleanPng's role to decode the header and create the
>   SkPngCodec
> - Make the interlaced PNG decoder report how many lines were
>   initialized during an incomplete decode
> - Make initializeSwizzler return a bool instead of an SkCodec::Result
>   (It only returned kSuccess or kInvalidInput anyway)
>
> SkIcoCodec:
> - Implement the new APIs; supported for PNG in ICO
>
> SkSampledCodec:
> - Call the new method for decoding scanlines, and fall back to the old
>   method if the new version is unimplemented
> - Remove references to kNone_SkScanlineOrder
>
> tests/CodecPartial:
> - Add a test which decodes part of an image, then finishes the decode,
>   and compares it to the straightforward method
>
> tests/CodecTest:
> - Add a test which decodes all scanlines using the new method
> - Repurpose the Codec_stripes test to decode using the new method in
>   sections rather than all at once
> - In the method check(), add a parameter for whether the image supports
>   the new method of scanline decoding, and be explicit about whether an
>   image supports incomplete
> - Test incomplete PNG decodes. We should have been doing it anyway for
>   non-interlaced (except for an image that is too small - one row), but
>   the new method supports interlaced incomplete as well
> - Make test_invalid_parameters test the new method
> - Add a test to ensure that it's safe to fall back to scanline decoding without
>   rewinding
>
> BUG=skia:4211
>
> The new version was generally faster than the old version (but not significantly so).
>
> Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/
>
> Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003
>
> Committed: https://skia.googlesource.com/skia/+/a4b09a117d4d1ba5dda372e6a2323e653766539e
>
> Committed: https://skia.googlesource.com/skia/+/30e78c9737ff4861dc4e3fa1e4cd010680ed6965

TBR=reed@google.com,msarett@google.com,scroggo@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4211

Review-Url: https://codereview.chromium.org/2026383002

8 years agoAdd SkPathPriv::IsSimpleClosedRect
bsalomon [Wed, 1 Jun 2016 16:42:31 +0000 (09:42 -0700)]
Add SkPathPriv::IsSimpleClosedRect

This function looks for "simple" rect paths. Simple here means begins and ends at a corner and is closed (either manually or with a close verb). Unlike SkPath::isRect this returns the starting point index (using the same start indexing scheme as SkPath::addRect).

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2017313002

Review-Url: https://codereview.chromium.org/2017313002

8 years agoAdd offset to memory allocations
jvanverth [Wed, 1 Jun 2016 16:39:15 +0000 (09:39 -0700)]
Add offset to memory allocations

This is setting up for suballocations within one large
allocation

BUG=skia:5031
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2018933004
TBR=bsalomon@google.com

Review-Url: https://codereview.chromium.org/2018933004

8 years agoRevert of Create SkColorSpaceXform to handle color conversions (patchset #10 id:26000...
msarett [Wed, 1 Jun 2016 16:16:04 +0000 (09:16 -0700)]
Revert of Create SkColorSpaceXform to handle color conversions (patchset #10 id:260001 of https://codereview.chromium.org/1952063002/ )

Reason for revert:
Google3 can't find qcms

Original issue's description:
> Create SkColorSpaceXform to handle color conversions
>
> Also adds testing of qcms color correction, so we can compare
> SkColorSpaceXform outputs to qcms outputs.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1952063002
>
> Committed: https://skia.googlesource.com/skia/+/740cc88ee3d63c75e52d31238f2a32600cc57a8c

TBR=reed@google.com,scroggo@google.com,mtklein@google.com,herb@google.com,brianosman@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review-Url: https://codereview.chromium.org/2023093004

8 years agoImplement support for rendering color emoji on Windows
kulshin [Wed, 1 Jun 2016 15:31:28 +0000 (08:31 -0700)]
Implement support for rendering color emoji on Windows

This change adds support to the DirectWrite scaler context
for detection of color fonts. If it detects a color font and
the glyph is a color glyph, it will use DirectWrite's
TranslateColorGlyphRun API and generate an image of the
glyph that can then be rendered.

Chromium tests: https://codereview.chromium.org/2003853002

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1984943002

Review-Url: https://codereview.chromium.org/1984943002

8 years agoCreate SkColorSpaceXform to handle color conversions
msarett [Wed, 1 Jun 2016 14:55:30 +0000 (07:55 -0700)]
Create SkColorSpaceXform to handle color conversions

Also adds testing of qcms color correction, so we can compare
SkColorSpaceXform outputs to qcms outputs.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1952063002

Review-Url: https://codereview.chromium.org/1952063002

8 years agoMake SkPngCodec decode progressively.
scroggo [Wed, 1 Jun 2016 14:31:28 +0000 (07:31 -0700)]
Make SkPngCodec decode progressively.

This is a step towards using SkCodec in Chromium, where progressive
decoding is necessary.

Switch from using png_read_row (which expects all the data to be
available) to png_process_data, which uses callbacks when rows are
available.

Create a new API for SkCodec, which supports progressive decoding and
scanline decoding. Future changes will switch the other clients off of
startScanlineDecode and get/skip-Scanlines to the new API.

Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced
PNG images. In the new API, interlaced PNG fits kTopDown. Also remove
updateCurrScanline(), which was only used by the old implementation for
interlaced PNG.

DMSrcSink:
- In CodecSrc::kScanline_Mode, use the new method for scanline decoding
for the supported formats (just PNG and PNG-in-ICO for now).

fuzz.cpp:
- Remove reference to kNone_ScanlineOrder

SkCodec:
- Add new APIs:
    - startIncrementalDecode
    - incrementalDecode
- Remove kNone_SkScanlineOrder and updateCurrScanline()

SkPngCodec:
- Implement new APIs
- Switch from sk_read_fn/png_read_row etc to png_process_data
- Expand AutoCleanPng's role to decode the header and create the
  SkPngCodec
- Make the interlaced PNG decoder report how many lines were
  initialized during an incomplete decode
- Make initializeSwizzler return a bool instead of an SkCodec::Result
  (It only returned kSuccess or kInvalidInput anyway)

SkIcoCodec:
- Implement the new APIs; supported for PNG in ICO

SkSampledCodec:
- Call the new method for decoding scanlines, and fall back to the old
  method if the new version is unimplemented
- Remove references to kNone_SkScanlineOrder

tests/CodecPartial:
- Add a test which decodes part of an image, then finishes the decode,
  and compares it to the straightforward method

tests/CodecTest:
- Add a test which decodes all scanlines using the new method
- Repurpose the Codec_stripes test to decode using the new method in
  sections rather than all at once
- In the method check(), add a parameter for whether the image supports
  the new method of scanline decoding, and be explicit about whether an
  image supports incomplete
- Test incomplete PNG decodes. We should have been doing it anyway for
  non-interlaced (except for an image that is too small - one row), but
  the new method supports interlaced incomplete as well
- Make test_invalid_parameters test the new method
- Add a test to ensure that it's safe to fall back to scanline decoding without
  rewinding

BUG=skia:4211

The new version was generally faster than the old version (but not significantly so).

Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/

Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003

Committed: https://skia.googlesource.com/skia/+/a4b09a117d4d1ba5dda372e6a2323e653766539e

Review-Url: https://codereview.chromium.org/1997703003

8 years ago[GN] Add support for disabling opts via SK_BUILD_NO_OPTS define.
sdefresne [Wed, 1 Jun 2016 14:08:56 +0000 (07:08 -0700)]
[GN] Add support for disabling opts via SK_BUILD_NO_OPTS define.

When targetting iOS and using gyp to generate the build files, it is not
possible to select files to build depending on the architecture. Due to
that, the skia code was disabling all optimisation when SK_BUILD_FOR_IOS
was defined.

Since it is possible to select the correct optimised version when using
gn, this pessimisation is hurting the build. Introduce a new define to
disable the optimisation SK_BUILD_NO_OPTS. It will be used by Chromium
when building skia for iOS with gyp but not gn.

Define SK_BUILD_NO_OPTS along-side SK_BUILD_FOR_IOS for all files that
look like build configuration (Xcode projects, gyp configuration files,
public.bzl) in order to avoid introducing breakage on those builds.

BUG=607933
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2002423002

Review-Url: https://codereview.chromium.org/2002423002

8 years agoalways compute a cubic tangent
caryclark [Wed, 1 Jun 2016 11:42:02 +0000 (04:42 -0700)]
always compute a cubic tangent

If the inflection of a cubic is also the cusp, the tangent
is degenerate at that t when computed directly. Rather than
giving up, subdivide the curve there and use the computed
control points to compute the tangent.

This strategy also removes the error paths where the tangent
formerly could not be computed.

R=reed@google.com
BUG=615686
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2020293002

Review-Url: https://codereview.chromium.org/2020293002

8 years agoCheck results from calls to SkCubicClipper::ChopMonoAtY.
mbarbella [Tue, 31 May 2016 21:44:01 +0000 (14:44 -0700)]
Check results from calls to SkCubicClipper::ChopMonoAtY.

BUG=613918
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2006143009

Review-Url: https://codereview.chromium.org/2006143009

8 years agoRevert of Make SkPngCodec decode progressively. (patchset #18 id:340001 of https...
scroggo [Tue, 31 May 2016 20:52:47 +0000 (13:52 -0700)]
Revert of Make SkPngCodec decode progressively. (patchset #18 id:340001 of https://codereview.chromium.org/1997703003/ )

Reason for revert:
This is failing tests and then crashing on Google3 [1]. The crashes are fixed by crrev.com/2026873002, but to fix the builder we'll need to upgrade its version of libpng.

[1] https://sponge.corp.google.com/target?id=e545ef55-4da4-4931-9524-1ac92acb61b1&target=//third_party/skia/HEAD:dm#shard=1|run=1|attempt=1|page=-1

Original issue's description:
> Make SkPngCodec decode progressively.
>
> This is a step towards using SkCodec in Chromium, where progressive
> decoding is necessary.
>
> Switch from using png_read_row (which expects all the data to be
> available) to png_process_data, which uses callbacks when rows are
> available.
>
> Create a new API for SkCodec, which supports progressive decoding and
> scanline decoding. Future changes will switch the other clients off of
> startScanlineDecode and get/skip-Scanlines to the new API.
>
> Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced
> PNG images. In the new API, interlaced PNG fits kTopDown. Also remove
> updateCurrScanline(), which was only used by the old implementation for
> interlaced PNG.
>
> DMSrcSink:
> - In CodecSrc::kScanline_Mode, use the new method for scanline decoding
> for the supported formats (just PNG and PNG-in-ICO for now).
>
> fuzz.cpp:
> - Remove reference to kNone_ScanlineOrder
>
> SkCodec:
> - Add new APIs:
>     - startIncrementalDecode
>     - incrementalDecode
> - Remove kNone_SkScanlineOrder and updateCurrScanline()
>
> SkPngCodec:
> - Implement new APIs
> - Switch from sk_read_fn/png_read_row etc to png_process_data
> - Expand AutoCleanPng's role to decode the header and create the
>   SkPngCodec
> - Make the interlaced PNG decoder report how many lines were
>   initialized during an incomplete decode
> - Make initializeSwizzler return a bool instead of an SkCodec::Result
>   (It only returned kSuccess or kInvalidInput anyway)
>
> SkIcoCodec:
> - Implement the new APIs; supported for PNG in ICO
>
> SkSampledCodec:
> - Call the new method for decoding scanlines, and fall back to the old
>   method if the new version is unimplemented
> - Remove references to kNone_SkScanlineOrder
>
> tests/CodecPartial:
> - Add a test which decodes part of an image, then finishes the decode,
>   and compares it to the straightforward method
>
> tests/CodecTest:
> - Add a test which decodes all scanlines using the new method
> - Repurpose the Codec_stripes test to decode using the new method in
>   sections rather than all at once
> - In the method check(), add a parameter for whether the image supports
>   the new method of scanline decoding, and be explicit about whether an
>   image supports incomplete
> - Test incomplete PNG decodes. We should have been doing it anyway for
>   non-interlaced (except for an image that is too small - one row), but
>   the new method supports interlaced incomplete as well
> - Make test_invalid_parameters test the new method
> - Add a test to ensure that it's safe to fall back to scanline decoding without
>   rewinding
>
> BUG=skia:4211
>
> The new version was generally faster than the old version (but not significantly so).
>
> Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/
>
> Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003
>
> Committed: https://skia.googlesource.com/skia/+/a4b09a117d4d1ba5dda372e6a2323e653766539e

TBR=reed@google.com,msarett@google.com,scroggo@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4211

Review-Url: https://codereview.chromium.org/2023103002

8 years agoskiaserve: Add /quitquitquit handler. Will be used in the hosted debugger.
jcgregorio [Tue, 31 May 2016 19:51:31 +0000 (12:51 -0700)]
skiaserve: Add /quitquitquit handler. Will be used in the hosted debugger.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2023023002
TBR=ethannicholas

Review-Url: https://codereview.chromium.org/2023023002

8 years agoMake use of new SkLights class
robertphillips [Tue, 31 May 2016 19:08:25 +0000 (12:08 -0700)]
Make use of new SkLights class

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2026763002

Review-Url: https://codereview.chromium.org/2026763002

8 years agoFix build error
scroggo [Tue, 31 May 2016 18:47:39 +0000 (11:47 -0700)]
Fix build error

TBR=herb@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2021393002

Review-Url: https://codereview.chromium.org/2021393002

8 years agoSkTypeface::MakeFromName to take SkFontStyle.
mboc [Tue, 31 May 2016 18:42:36 +0000 (11:42 -0700)]
SkTypeface::MakeFromName to take SkFontStyle.

SkTypeface::MakeFromName currently takes SkTypeface::Style,
which is quite limited. This starts the transition to this
function taking SkFontStyle instead.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1818043002

TBR=reed
He said it sounded like a good idea.

Review-Url: https://codereview.chromium.org/1818043002

8 years agoSkXPS: clean up SkConstexprMath
halcanary [Tue, 31 May 2016 18:23:42 +0000 (11:23 -0700)]
SkXPS: clean up SkConstexprMath

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2000853003

Review-Url: https://codereview.chromium.org/2000853003

8 years agoSkEncodedInfo: simplify
halcanary [Tue, 31 May 2016 17:57:36 +0000 (10:57 -0700)]
SkEncodedInfo: simplify

No public API change.
TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2025713003

Review-Url: https://codereview.chromium.org/2025713003

8 years agoFix bug where SkPath will convert an arc to an oval and change the starting point.
bsalomon [Tue, 31 May 2016 17:42:16 +0000 (10:42 -0700)]
Fix bug where SkPath will convert an arc to an oval and change the starting point.

Adds unit tests to test the behavior of addArc to oval conversions.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2017743002

Review-Url: https://codereview.chromium.org/2017743002

8 years agoUse special case for 0x00 and 0xFF alpha to go faster.
herb [Tue, 31 May 2016 17:04:24 +0000 (10:04 -0700)]
Use special case for 0x00 and 0xFF alpha to go faster.
Base          Exp           Ratio           Name
3916010   3077544   0.785888  top25desk_amazon_com.skp_1
3462512   2776580   0.801898  top25desk_google_com_calendar_.skp_1
3446330   3134187   0.909427  top25desk_ebay_com.skp_1
590474    546375    0.925316  top25desk_techcrunch_com.skp_1
3804991   3544162   0.931451  top25desk_google_com__hl_en_q_b.skp_1
996037    939960    0.9437    top25desk_blogger.skp_1
973264    922677    0.948023  top25desk_wikipedia__1_tab_.skp_1
4514050   4295660   0.95162   top25desk_docs___1_open_documen.skp_1
4255383   4110057   0.965849  top25desk_linkedin.skp_1
8408717   8191843   0.974208  top25desk_booking_com.skp_1
2878529   2806501   0.974977  top25desk_plus_google_com_11003.skp_1
11509894  11254486  0.97781   top25desk_pinterest.skp_1
9132514   9010635   0.986654  top25desk_weather_com.skp_1
6504720   6419592   0.986913  top25desk_sports_yahoo_com_.skp_1
9136774   9033870   0.988737  top25desk_answers_yahoo_com.skp_1
4836199   4799784   0.99247   top25desk_news_yahoo_com.skp_1
1393650   1384065   0.993122  top25desk_games_yahoo_com.skp_1
6779678   6735278   0.993451  top25desk_youtube_com.skp_1
10926943  10882308  0.995915  top25desk_espn.skp_1
4259514   4245489   0.996707  top25desk_facebook.skp_1
10955293  10947657  0.999303  top25desk_google_com_search_q_c.skp_1
9153575   9207386   1.00588   top25desk_twitter.skp_1
3865942   3906345   1.01045   top25desk_wordpress.skp_1
4180009   4305530   1.03003   top25desk_mail_google_com_mail_.skp_1

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2020463003

Review-Url: https://codereview.chromium.org/2020463003

8 years agoAdd mip support to *DeferredTextureImageData functions.
cblume [Tue, 31 May 2016 16:55:08 +0000 (09:55 -0700)]
Add mip support to *DeferredTextureImageData functions.

As a first step, I want to add the ability to store multiple mip levels
in the opaque object.

I will follow this with using SkMipMap to generate and store the mip
levels and generating the texture from those stored mip levels.

R=bsalomon@chromium.org
CC=ericrk@chromium.org
BUG=578304
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2023573002

Review-Url: https://codereview.chromium.org/2023573002

8 years agoFilter out degenerate contours in GrConvexPolyEffect
lsalzman [Tue, 31 May 2016 16:46:00 +0000 (09:46 -0700)]
Filter out degenerate contours in GrConvexPolyEffect

As noticed in a downstream Firefox bug report https://bugzilla.mozilla.org/show_bug.cgi?id=1255062

If a path such as (moveTo, moveTo, lineTo, lineTo, close) is supplied,
and if the non-degenerate contour is convex, the convexity test will
pass, and GrConvexPolyEffect will be used.

However, the path's raw points are used to build the edge list, which
does not filter out degenerate contours. This may cause the polygon to
fail to draw.

This patch ensures that the degenerate contours are filtered out by
using an iterator as the path convexity test does.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2018613003

Review-Url: https://codereview.chromium.org/2018613003

8 years agoAdd missing files for GL unix viewer.
jvanverth [Tue, 31 May 2016 16:39:20 +0000 (09:39 -0700)]
Add missing files for GL unix viewer.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2023943004

Review-Url: https://codereview.chromium.org/2023943004

8 years agoFix integer overflow in SkColorSpace
msarett [Tue, 31 May 2016 16:13:04 +0000 (09:13 -0700)]
Fix integer overflow in SkColorSpace

BUG=skia:5361
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2020303002

Review-Url: https://codereview.chromium.org/2020303002

8 years agoAdd new GM to exercise stroked line special case
robertphillips [Tue, 31 May 2016 13:59:18 +0000 (06:59 -0700)]
Add new GM to exercise stroked line special case

This is a companion to: https://codereview.chromium.org/2023693002/ (Handle stroked single line special case in Ganesh)

The idea is to land this CL first so any regressions are visible.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2019193002

Review-Url: https://codereview.chromium.org/2019193002

8 years agoAdd SkLights class
robertphillips [Tue, 31 May 2016 13:56:28 +0000 (06:56 -0700)]
Add SkLights class

The SkLights class is intended to be passed to the drawLitAtlas call (and then on to SkLightingShader).

This CL is calved off of https://codereview.chromium.org/1261433009/ (Add SkCanvas::drawLitAtlas call).
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1255133004

Review-Url: https://codereview.chromium.org/1255133004

8 years agoRemove unused SkPictureGpuAnalyzer::analyze()
fmalita [Tue, 31 May 2016 13:32:57 +0000 (06:32 -0700)]
Remove unused SkPictureGpuAnalyzer::analyze()

After https://crrev.com/c64bd21780da13051846ae45b54a1e6c6bf6ac52,
Chromium no longer uses analyze().

R=mtklein@google.com,reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2014983002

Review-Url: https://codereview.chromium.org/2014983002

8 years agoMake SkPngCodec decode progressively.
scroggo [Tue, 31 May 2016 13:30:14 +0000 (06:30 -0700)]
Make SkPngCodec decode progressively.

This is a step towards using SkCodec in Chromium, where progressive
decoding is necessary.

Switch from using png_read_row (which expects all the data to be
available) to png_process_data, which uses callbacks when rows are
available.

Create a new API for SkCodec, which supports progressive decoding and
scanline decoding. Future changes will switch the other clients off of
startScanlineDecode and get/skip-Scanlines to the new API.

Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced
PNG images. In the new API, interlaced PNG fits kTopDown. Also remove
updateCurrScanline(), which was only used by the old implementation for
interlaced PNG.

DMSrcSink:
- In CodecSrc::kScanline_Mode, use the new method for scanline decoding
for the supported formats (just PNG and PNG-in-ICO for now).

fuzz.cpp:
- Remove reference to kNone_ScanlineOrder

SkCodec:
- Add new APIs:
    - startIncrementalDecode
    - incrementalDecode
- Remove kNone_SkScanlineOrder and updateCurrScanline()

SkPngCodec:
- Implement new APIs
- Switch from sk_read_fn/png_read_row etc to png_process_data
- Expand AutoCleanPng's role to decode the header and create the
  SkPngCodec
- Make the interlaced PNG decoder report how many lines were
  initialized during an incomplete decode
- Make initializeSwizzler return a bool instead of an SkCodec::Result
  (It only returned kSuccess or kInvalidInput anyway)

SkIcoCodec:
- Implement the new APIs; supported for PNG in ICO

SkSampledCodec:
- Call the new method for decoding scanlines, and fall back to the old
  method if the new version is unimplemented
- Remove references to kNone_SkScanlineOrder

tests/CodecPartial:
- Add a test which decodes part of an image, then finishes the decode,
  and compares it to the straightforward method

tests/CodecTest:
- Add a test which decodes all scanlines using the new method
- Repurpose the Codec_stripes test to decode using the new method in
  sections rather than all at once
- In the method check(), add a parameter for whether the image supports
  the new method of scanline decoding, and be explicit about whether an
  image supports incomplete
- Test incomplete PNG decodes. We should have been doing it anyway for
  non-interlaced (except for an image that is too small - one row), but
  the new method supports interlaced incomplete as well
- Make test_invalid_parameters test the new method
- Add a test to ensure that it's safe to fall back to scanline decoding without
  rewinding

BUG=skia:4211

The new version was generally faster than the old version (but not significantly so).

Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/

Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003

Review-Url: https://codereview.chromium.org/1997703003

8 years agoFix incorrect sampler locations when using CHROMIUM_bind_uniform_location
kkinnunen [Tue, 31 May 2016 05:30:24 +0000 (22:30 -0700)]
Fix incorrect sampler locations when using CHROMIUM_bind_uniform_location

Fix incorrect sampler locations when using
CHROMIUM_bind_uniform_location.

The extension is not used at the moment, though.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2023783002

Review-Url: https://codereview.chromium.org/2023783002

8 years agoUpdate SKP version
update-skps [Sun, 29 May 2016 08:00:14 +0000 (01:00 -0700)]
Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=
NO_MERGE_BUILDS
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2024513002

Review-Url: https://codereview.chromium.org/2024513002

8 years agoChange parameters to GrPipelineBuilder's ctor
robertphillips [Sat, 28 May 2016 15:51:06 +0000 (08:51 -0700)]
Change parameters to GrPipelineBuilder's ctor

This is just plumbing prep to remove the GrRenderTarget from the GrPipelineBuilder.

Split out of: https://codereview.chromium.org/1988923002/ (Declassify GrClipMaskManager and Remove GrRenderTarget and GrDrawTarget from GrPipelineBuilder)

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2015333002

Review-Url: https://codereview.chromium.org/2015333002

8 years agoAdd interesting features to visualize_color_gamut
msarett [Fri, 27 May 2016 21:01:02 +0000 (14:01 -0700)]
Add interesting features to visualize_color_gamut

(1) Draw triangles onto a D50 color gamut so they are
    actually in the correct place.
(2) Add options to display canonical sRGB and Adobe RGB
    gamuts.
(3) Label the R, G, and B points of the color gamut.
(4) Decode and reencode the input image without color
    correction, so we can compare to the corrected version.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2017153002

Review-Url: https://codereview.chromium.org/2017153002

8 years agoUpdate VC++ version in Windows quick start.
bungeman [Fri, 27 May 2016 19:27:33 +0000 (12:27 -0700)]
Update VC++ version in Windows quick start.

NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=2019093002

Review-Url: https://codereview.chromium.org/2019093002

8 years agoRewrite assignment operator on SkBitmap, fix asan
msarett [Fri, 27 May 2016 18:31:55 +0000 (11:31 -0700)]
Rewrite assignment operator on SkBitmap, fix asan

TBR=mtklein@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2020683002

Review-Url: https://codereview.chromium.org/2020683002

8 years agoisolate_android_sdk: Find or download isolate binary, fix infra path
borenet [Fri, 27 May 2016 18:11:22 +0000 (11:11 -0700)]
isolate_android_sdk: Find or download isolate binary, fix infra path

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2016353003

Review-Url: https://codereview.chromium.org/2016353003

8 years agoMove forceSRGB to SkCommonFlags.
herb [Fri, 27 May 2016 17:57:58 +0000 (10:57 -0700)]
Move forceSRGB to SkCommonFlags.

This fixes a duplicate symbol problem with the iOS build.

Add common forceSRGB flag.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2019073002

Review-Url: https://codereview.chromium.org/2019073002

8 years agoClean up SkFloatBits
mtklein [Fri, 27 May 2016 17:47:31 +0000 (10:47 -0700)]
Clean up SkFloatBits

- remove dead code
- rewrite float -> int converters

The strategy for the new converters is:
   - convert input to double
   - floor/ceil/round in double space
   - pin that double to [SK_MinS32, SK_MaxS32]
   - truncate that double to int32_t

This simpler strategy does not work:
   - floor/ceil/round in float space
   - pin that float to [SK_MinS32, SK_MaxS32]
   - truncate that float to int32_t

SK_MinS32 and SK_MaxS32 are not representable as floats:
they round to the nearest float, ±2^31, which makes the
pin insufficient for floats near SK_MinS32 (-2^31+1) or
SK_MaxS32 (+2^31-1).

float only has 24 bits of precision, and we need 31.
double can represent all integers up to 50-something bits.

An alternative is to pin in float to ±2147483520, the last
exactly representable float before SK_MaxS32 (127 too small).

Our tests test that we round as floor(x+0.5), which can
return different numbers than round(x) for negative x.
So this CL explicitly uses floor(x+0.5).

I've updated the tests with ±inf and ±NaN, and tried to
make them a little clearer, especially using SK_MinS32
instead of -SK_MaxS32.

I have not timed anything here.  I have never seen any of these
methods in a profile.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2012333003

Review-Url: https://codereview.chromium.org/2012333003

8 years agoMake SkPath::isOval() and SkPath::isRRect return the orientation and starting index.
bsalomon [Fri, 27 May 2016 16:17:04 +0000 (09:17 -0700)]
Make SkPath::isOval() and SkPath::isRRect return the orientation and starting index.

These are tracked in SkPathRef.

Unit tests are updated to test that the returned values are correct.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2012233002

Review-Url: https://codereview.chromium.org/2012233002

8 years agoNow we can use drawer to view the state information of the native app, and set its...
liyuqian [Fri, 27 May 2016 15:52:52 +0000 (08:52 -0700)]
Now we can use drawer to view the state information of the native app, and set its state using the spinner.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2004633002

Committed: https://skia.googlesource.com/skia/+/4e4e30823fba0047b93a56bfcf05c04ca57e82ec

Review-Url: https://codereview.chromium.org/2004633002

8 years agoAdd forceSRGB flag to dm.
herb [Fri, 27 May 2016 15:15:05 +0000 (08:15 -0700)]
Add forceSRGB flag to dm.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2011333002

Review-Url: https://codereview.chromium.org/2011333002

8 years agoUpdate Android SDK
liyuqian [Fri, 27 May 2016 15:08:34 +0000 (08:08 -0700)]
Update Android SDK

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2016363002

Review-Url: https://codereview.chromium.org/2016363002

8 years agoAdd SkColorSpace to SkImageInfo
msarett [Fri, 27 May 2016 14:39:02 +0000 (07:39 -0700)]
Add SkColorSpace to SkImageInfo

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2000713003

Review-Url: https://codereview.chromium.org/2000713003

8 years agoAdd OpenGL support to Linux viewer
jvanverth [Fri, 27 May 2016 13:47:08 +0000 (06:47 -0700)]
Add OpenGL support to Linux viewer

Also adds a command line option (--vulkan) to choose between
Vulkan and OpenGL.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2011473003

Committed: https://skia.googlesource.com/skia/+/10b3815a11c39b154ec7b74a2af43e019b50d48c

Review-Url: https://codereview.chromium.org/2011473003

8 years agoRevert of Add drawer with state information (patchset #8 id:140001 of https://coderev...
kjlubick [Fri, 27 May 2016 13:22:01 +0000 (06:22 -0700)]
Revert of Add drawer with state information (patchset #8 id:140001 of https://codereview.chromium.org/2004633002/ )

Reason for revert:
Broke Build-Ubuntu-GCC-Arm7-Debug-Android_Vulkan.  See https://uberchromegw.corp.google.com/i/client.skia.compile/builders/Build-Ubuntu-GCC-Arm7-Debug-Android_Vulkan/builds/134/steps/compile_skia%20on%20Ubuntu/logs/stdio.

Original issue's description:
> Now we can use drawer to view the state information of the native app, and set its state using the spinner.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2004633002
>
> Committed: https://skia.googlesource.com/skia/+/4e4e30823fba0047b93a56bfcf05c04ca57e82ec

TBR=djsollen@google.com,jvanverth@google.com,liyuqian@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review-Url: https://codereview.chromium.org/2016343002

8 years agoRevert of Add OpenGL support to Linux viewer (patchset #5 id:80001 of https://coderev...
kjlubick [Fri, 27 May 2016 13:20:10 +0000 (06:20 -0700)]
Revert of Add OpenGL support to Linux viewer (patchset #5 id:80001 of https://codereview.chromium.org/2011473003/ )

Reason for revert:
Build-Ubuntu-GCC-Arm7-Debug-Android_Vulkan is broken and this patch prevents the revert of the culprit

Original issue's description:
> Add OpenGL support to Linux viewer
>
> Also adds a command line option (--vulkan) to choose between
> Vulkan and OpenGL.
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2011473003
>
> Committed: https://skia.googlesource.com/skia/+/10b3815a11c39b154ec7b74a2af43e019b50d48c

TBR=egdaniel@google.com,bsalomon@google.com,brianosman@google.com,jvanverth@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review-Url: https://codereview.chromium.org/2015213002

8 years agoAdd OpenGL support to Linux viewer
jvanverth [Fri, 27 May 2016 12:47:00 +0000 (05:47 -0700)]
Add OpenGL support to Linux viewer

Also adds a command line option (--vulkan) to choose between
Vulkan and OpenGL.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2011473003

Review-Url: https://codereview.chromium.org/2011473003

8 years agoRemove VisualBench and its Android implementation.
djsollen [Fri, 27 May 2016 12:39:29 +0000 (05:39 -0700)]
Remove VisualBench and its Android implementation.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2018603003

Review-Url: https://codereview.chromium.org/2018603003

8 years agopin before calling acos
caryclark [Fri, 27 May 2016 12:24:37 +0000 (05:24 -0700)]
pin before calling acos

Adobe reports some user crashes in acos(). While the cause is
unknown, it's safe and may help stability to pin the input
in case the arguments drifted slightly outside [-1, 1].

R=reed@google.com
BUG=skia:5222
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2006653006

Review-Url: https://codereview.chromium.org/2006653006

8 years agotoy to play with antialias raytracing
caryclark [Fri, 27 May 2016 12:13:26 +0000 (05:13 -0700)]
toy to play with antialias raytracing

Here's the ray tracer I wrote.
I've commented out the calls to experimental
stroking -- enough of it remains working
to help visualize what you've been talking
to me about.

R=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2014993002

Review-Url: https://codereview.chromium.org/2014993002

8 years agoSilence ASAN int32 overflow warning
fmalita [Fri, 27 May 2016 01:10:24 +0000 (18:10 -0700)]
Silence ASAN int32 overflow warning

It's fine to overflow SK_MaxS32 by one, the subsequent cast ensures
correct clamping.  But we need to cast earlier in order to make ASAN
happy.

TBR=mtklein@google.com,reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2013243002

Review-Url: https://codereview.chromium.org/2013243002

8 years agoadd --forceSRGB option to nanobench, to get all src images to be SRGB
reed [Thu, 26 May 2016 21:06:43 +0000 (14:06 -0700)]
add --forceSRGB option to nanobench, to get all src images to be SRGB

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2008333002

Review-Url: https://codereview.chromium.org/2008333002

8 years agoNow we can use drawer to view the state information of the native app, and set its...
liyuqian [Thu, 26 May 2016 19:43:43 +0000 (12:43 -0700)]
Now we can use drawer to view the state information of the native app, and set its state using the spinner.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2004633002

Review-Url: https://codereview.chromium.org/2004633002

8 years agoadd more SK_WARN_UNUSED_RESULT attributes to functions that ignore out-params if...
reed [Thu, 26 May 2016 19:13:26 +0000 (12:13 -0700)]
add more SK_WARN_UNUSED_RESULT attributes to functions that ignore out-params if they fail

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2012943003

TBR=

Review-Url: https://codereview.chromium.org/2012943003

8 years agoTemporarily change GalaxyS3 product.board
borenet [Thu, 26 May 2016 18:21:02 +0000 (11:21 -0700)]
Temporarily change GalaxyS3 product.board

Swarming detects this as "m0" for some reason. Change our requested
dimension until this is fixed.

BUG=skia:5185
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2018513004

Review-Url: https://codereview.chromium.org/2018513004

8 years agoFix int32 overflow in LinearGradientContext::shade4_dx_clamp
fmalita [Thu, 26 May 2016 18:13:52 +0000 (11:13 -0700)]
Fix int32 overflow in LinearGradientContext::shade4_dx_clamp

The unconditional increment in shade4_dx_clamp can overflow int32

=> n == SK_MinS32
=> count ~= SK_MinS32
=> we skip the main shader loop 'cause count < 0

R=reed@google.com,mtklein@google.com
BUG=chromium:599458
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2010843002

Review-Url: https://codereview.chromium.org/2010843002

8 years agoAdd exact version of qcms used by Chrome for testing and comparison
msarett [Thu, 26 May 2016 18:05:37 +0000 (11:05 -0700)]
Add exact version of qcms used by Chrome for testing and comparison

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2014023003

Review-Url: https://codereview.chromium.org/2014023003

8 years agoSrcover for sprite blitters.
herb [Thu, 26 May 2016 17:56:17 +0000 (10:56 -0700)]
Srcover for sprite blitters.

In order for this code to run, the gDefaultProfileIsSRGB flag needs to be true.

min base     min exp    percent   name
5601856   4689911   0.837207  top25desk_espn.skp_1
3491515   3202806   0.917311  top25desk_facebook.skp_1
5110247   4865740   0.952154  top25desk_weather_com.skp_1
605445    585520    0.96709   top25desk_techcrunch_com.skp_1
1007151   986193    0.979191  top25desk_wikipedia__1_tab_.skp_1
5951286   5889979   0.989699  top25desk_sports_yahoo_com_.skp_1
2825583   2804853   0.992663  top25desk_plus_google_com_11003.skp_1
8839265   8823249   0.998188  top25desk_twitter.skp_1
4169125   4168882   0.999942  top25desk_docs___1_open_documen.skp_1
6615327   6620663   1.00081   top25desk_youtube_com.skp_1
4613903   4647583   1.0073    top25desk_wordpress.skp_1
2532280   2554154   1.00864   top25desk_ebay_com.skp_1
4015689   4063584   1.01193   top25desk_google_com__hl_en_q_b.skp_1
9427478   9579203   1.01609   top25desk_answers_yahoo_com.skp_1
7403901   7542770   1.01876   top25desk_booking_com.skp_1
12249953  12528353  1.02273   top25desk_google_com_search_q_c.skp_1
1078648   1111050   1.03004   top25desk_games_yahoo_com.skp_1
7232627   7481555   1.03442   top25desk_pinterest.skp_1
2996819   3112091   1.03846   top25desk_google_com_calendar_.skp_1
2181531   2271677   1.04132   top25desk_amazon_com.skp_1
925245    987545    1.06733   top25desk_blogger.skp_1
4143359   4442607   1.07222   top25desk_linkedin.skp_1
4370962   4744580   1.08548   top25desk_news_yahoo_com.skp_1
4284025   4735094   1.10529   top25desk_mail_google_com_mail_.skp_1

[mtklein] We measured the noise here to be [-5%, +8%], so most of these changes fall within the noise.  We manually confirmed the two above that noise window (yahoo and mail) are also noise... srcover_srgb_srgb() did not figure prominently in their profiles.  The espn and facebook improvements look real.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1996683003

Review-Url: https://codereview.chromium.org/1996683003

8 years agofix security bug
caryclark [Thu, 26 May 2016 16:01:47 +0000 (09:01 -0700)]
fix security bug

This fix is a tradeoff. It changes intersection to
treat a case where one coincident run is intersected at one point
and the other edge is not as continuing to be a span.

The old code tried to treat this as a single point.
The old code is probably right, but this change alone
made the data structures inconsistent. Later, extending
the coincident runs would fail by incorrectly discarding
the single point intersection.

As a result, this fixes the security test and one other, but
makes a different test fail. Isolating the failure uncovered
a reduced case that fails with and without the change, so
there are more serious problems here. Those problems are
addressed in a separate CL.

Many of the test edits below remove ill-thought out debugging
messaging that fire off global state, which isn't usable
in a multi-threaded test environment.

In the end, with this fix, all existing tests (modulo one
new failure and one new non-failure) pass in debug and
in the extended release test suites.

TBR=reed@google.com
BUG=614248
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2018513003

Review-Url: https://codereview.chromium.org/2018513003

8 years agoSkPDF/SkColorSpace: use writeToICC
halcanary [Thu, 26 May 2016 15:31:06 +0000 (08:31 -0700)]
SkPDF/SkColorSpace: use writeToICC

.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2012133003

Review-Url: https://codereview.chromium.org/2012133003

8 years agoComplex clipPath accounting
fmalita [Wed, 25 May 2016 18:31:04 +0000 (11:31 -0700)]
Complex clipPath accounting

Add a clipPath heuristic to SkPathCounter, and extend
SkPictureGpuAnalyzer to support external clipPath() op accounting.

BUG=skia:5347
R=reed@google.com,mtklein@google.com,senorblanco@chromium.org,bsalomon@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2000423005

Review-Url: https://codereview.chromium.org/2000423005

8 years agoAssign sRGB color space to unmarked jpegs and pngs
msarett [Wed, 25 May 2016 17:13:53 +0000 (10:13 -0700)]
Assign sRGB color space to unmarked jpegs and pngs

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2009813003

Review-Url: https://codereview.chromium.org/2009813003

8 years agoWrite ICC profiles from SkColorSpace object
msarett [Wed, 25 May 2016 15:53:40 +0000 (08:53 -0700)]
Write ICC profiles from SkColorSpace object

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2001203003

Review-Url: https://codereview.chromium.org/2001203003

8 years agoDon't store resources with a unique key in GrResourceCache's fScratchMap
robertphillips [Tue, 24 May 2016 21:17:12 +0000 (14:17 -0700)]
Don't store resources with a unique key in GrResourceCache's fScratchMap

The reasoning here is that resources with a unique key are never selected from fScratchMap and just clog up the search for an available resource.

This knocks a 200x loop over the SVGbouncingrects case from 264ms down to 164ms.

BUG=603969
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2008083002

Review-Url: https://codereview.chromium.org/2008083002

8 years agoMake Sk32ToBool inline again to silence compiler warning.
bsalomon [Tue, 24 May 2016 20:23:56 +0000 (13:23 -0700)]
Make Sk32ToBool inline again to silence compiler warning.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2007753003

Review-Url: https://codereview.chromium.org/2007753003

8 years agoCMake: skia_link_arguments.txt repects BUILD_SHARED_LIBS
halcanary [Tue, 24 May 2016 18:23:23 +0000 (11:23 -0700)]
CMake: skia_link_arguments.txt repects BUILD_SHARED_LIBS

.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2007093002
CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Ubuntu-GCC-x86_64-Release-CMake-Trybot,Build-Mac-Clang-x86_64-Release-CMake-Trybot

Review-Url: https://codereview.chromium.org/2007093002

8 years agoSkColorSpace tweaks
msarett [Tue, 24 May 2016 17:16:53 +0000 (10:16 -0700)]
SkColorSpace tweaks

***Move includes to the correct file
***Add "const"
***Use std::move for more efficient refs and unrefs

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2005263002

Review-Url: https://codereview.chromium.org/2005263002

8 years agoCMake: control static/shared the normal CMake way.
mtklein [Tue, 24 May 2016 16:44:05 +0000 (09:44 -0700)]
CMake: control static/shared the normal CMake way.

This flips the default build mode to create a static libskia.

To create a shared libskia, pass -DBUILD_SHARED_LIBS=1 when running cmake.

BUG=skia:5341
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2009503002
CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Ubuntu-GCC-x86_64-Release-CMake-Trybot,Build-Mac-Clang-x86_64-Release-CMake-Trybot

Review-Url: https://codereview.chromium.org/2009503002

8 years agoUse nullptr instead of 0 in call to FontConfig.
bungeman [Tue, 24 May 2016 15:08:20 +0000 (08:08 -0700)]
Use nullptr instead of 0 in call to FontConfig.

Review-Url: https://codereview.chromium.org/2000253005

8 years agopthread_once -> SkOnce
mtklein [Tue, 24 May 2016 14:01:48 +0000 (07:01 -0700)]
pthread_once -> SkOnce

one fewer sync primitive...

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2003163004

Review-Url: https://codereview.chromium.org/2003163004