Tune linear->sRGB constants to round-trip all bytes.

I basically just ran a big 5-deep for-loop over the five constants here.
This is the first set of coefficients I found that round trips all bytes.
I suspect there are many such sets.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2162063003
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review-Url: https://codereview.chromium.org/2162063003

Refactor parsing and storage of SkGammas

Benefits:
(1) Parses and stores gamma tags in a single allocation.
(2) Recognizes equal gamma tags to skip parsing work and
    save memory.

Non-Benefits:
(1) Not less complicated.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2117773002

Review-Url: https://codereview.chromium.org/2117773002

re-disable save/restore peephole optimization

BUG=skia:5548, 629408
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2163283002

Review-Url: https://codereview.chromium.org/2163283002

SkLS now accepting nullptr for diffuse shader and normal source, now accurately handling alpha

This CL's base is the CL for taking in a diffuse shader into SkLS on the API side: https://codereview.chromium.org/2064153002

BUG=skia:5502,skia:5517
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2132113002

Review-Url: https://codereview.chromium.org/2132113002

Add vulkan sdk to CIPD

BUG=skia:5427
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2164953002

Review-Url: https://codereview.chromium.org/2164953002

re-chop if we fail on a big-bad-cubic

BUG=629455
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2159223005

Review-Url: https://codereview.chromium.org/2159223005

Improve assert reporting.

Make SkASSERTF output readable.
Ensure the assert predicate is stringified once.
Make the abort code consistent.

TBR=reed
This doesn't change any public API, most of this should be privatized.

Review-Url: https://codereview.chromium.org/2161103002

Remove SkDevice::accessRenderTarget virtual
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2167723002

Review-Url: https://codereview.chromium.org/2167723002

Add GrVkRecycledResource subclass

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2167453003

Review-Url: https://codereview.chromium.org/2167453003

Add integer array uniforms to GL and Vulkan

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2163173005

Review-Url: https://codereview.chromium.org/2163173005

Consolidate special case shape transformation logic in GrShapeTest.

Enable all tests on all geometry types.

Add conversion of fill+miter-stroke->fill for rect geometries in GrShape.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2157013003

Review-Url: https://codereview.chromium.org/2157013003

Add iOS to platform_os_emoji.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2164783004

Review-Url: https://codereview.chromium.org/2164783004

Fix context creation problems in dm

Fixes these issues:

1. The glXCreateContextAttribsARB call seems to modify the
context_attrib parameter upon failure. Since context_attrib was static,
if it first tried to create an OpenGL 4.4 context and it failed, further
attempts to create a context with lower versions would also fail.

2. Getting an OpenGL 3.0 context was never tried.

3. Context creation for ES 3.0 was failing on my machine. Since ES 2.0
looks like the intended version from the surrounding code, I set it to
that. It could be rewritten to use a similar loop as standard OpenGL and
try from the highest version to the lowest.

BUG=skia:5403
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2147513002

Review-Url: https://codereview.chromium.org/2147513002

use special-image for imagefilters and save/restore layer

add special virtuals to device, in preparation for using them instead of bitmap for imagefilters

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2155933002

patch from issue 2155933002 at patchset 20001 (http://crrev.com/2155933002#ps20001)

use specialimages instead of bitmaps for imagefiltering

Review-Url: https://codereview.chromium.org/2155063002

Add "Infra" to the builder_name_schema

BUG=skia:4763
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2154433002

Review-Url: https://codereview.chromium.org/2154433002

Fix color order for 565,4444, and I8 samplers.

10 GMs fixed on Linux/Win!

Also, don't unpremul I8 colors.  It's inconsistent with the other formats.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2166623002

Review-Url: https://codereview.chromium.org/2166623002

Port FuzzPathop from chromium

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2148023002

Review-Url: https://codereview.chromium.org/2148023002

pre-land special methods on device

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2161233002

Review-Url: https://codereview.chromium.org/2161233002

Stub in Mac viewer code.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2164683002

patch from issue 2164683002 at patchset 1 (http://crrev.com/2164683002#ps1)

Review-Url: https://codereview.chromium.org/2167503002

Remove unused MakeFromTextureCopy API

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2165803002

Review-Url: https://codereview.chromium.org/2165803002

Minor cleanup of GrStencilAndCoverTextContext

Split out of https://codereview.chromium.org/2163483002/ (Use SkFont in GrStencilAndCoverTextContext)

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2165673002

Review-Url: https://codereview.chromium.org/2165673002

Update rounding/clamping comments on sk_linear_to_srgb().

Rounding is enough.  No need for an explicit clamp if the inputs are in range.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2161223002

Review-Url: https://codereview.chromium.org/2161223002

Make SkOTTable_name.cpp UTF-8.

When editing on Windows this file became iso-8859-1.

Fix master-skia build

TBR=djsollen@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2159223003

Review-Url: https://codereview.chromium.org/2159223003

Fix color order on LCD text when using sRGB software backend.

BUG=skia:5182
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2166533002

Review-Url: https://codereview.chromium.org/2166533002

exclude test that fails on google3 asan

R=dogben@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2164703002

Review-Url: https://codereview.chromium.org/2164703002

fix fuzzer

Previous spans always have a valid next pointer. The final span does not.
Change the test for a valid link to take into consideration whether
the links are chased forwards or backwards.

TBR=reed@google.com
BUG=629454
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2166543002

Review-Url: https://codereview.chromium.org/2166543002

Disable qcms on build for Android framework

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2160863003

Review-Url: https://codereview.chromium.org/2160863003

Enable libjpeg-turbo features

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2161843002

Review-Url: https://codereview.chromium.org/2161843002

Clarify GrGradientEffect key enum

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2158323003

Review-Url: https://codereview.chromium.org/2158323003

Make sk_app::WindowContext directly create a SkSurface without an intermediate GrRenderTarget
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2165703002

Review-Url: https://codereview.chromium.org/2165703002

Improve naive SkColorXform to half floats

This should give us a good baseline to explore using SkRasterPipeline.

A particular colorxform to half float drops from 425us to 282us on my desktop.

Color Xform to Half Float (HP z620)
Original                              425us
Trans16 (not 32)                      355us
Vector Trans16                        378us
Trans16 + Keep Halfs in Vector        335us
Vector Trans16 + Keep Halfs in Vector 282us
Final                                 282us

Color Xform to Half Float (Nexus 5X)
Original                              556us
Final                                 472us

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2159993003
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review-Url: https://codereview.chromium.org/2159993003

Create DCs instead of textures in SRGBMipMaps test
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2162973002

Review-Url: https://codereview.chromium.org/2162973002

Create DCs rather than textures in GrConfigConversionEffect
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2162953002

Review-Url: https://codereview.chromium.org/2162953002

Simplify random_draw_context used by GLProgramsTest

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2165683002

Review-Url: https://codereview.chromium.org/2165683002

remove unused MakeFromPixmap

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2158913004

TBR=robertphilips@google.com

Review-Url: https://codereview.chromium.org/2158913004

GN: refactor third_party/gni

  This fixes the build on Linux (dep on third_party:zlib -> third_party/zlib).

  I've moved declare_args() {} back to each .gn file... seems like args want
  to be as local as possible in GN land.

  Additionally, refactor all the common third_party config and warning flag
  changes into a template, third_party.  This lets it all live together in a
  .gni: at head unwanted_configs can be in a .gni (it's just a variable) but
  config("no_warnings") (and thus third_party_configs) cannot, as configs
  cannot be part of .gni files.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2163653002

Review-Url: https://codereview.chromium.org/2163653002

Start from DC rather than texture in GrTexutreParamsAdjuster copy code.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2158383002

Review-Url: https://codereview.chromium.org/2158383002

Fix debugging code for tracing texture creation

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2164623002

Review-Url: https://codereview.chromium.org/2164623002

Make SkFont a bit more useable

Split out of https://codereview.chromium.org/2163483002/ (Use SkFont in GrStencilAndCoverTextContext)

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2163633002

Review-Url: https://codereview.chromium.org/2163633002

Convert readSurfacePixels to use a draw context instead of a texture
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2165523003

Review-Url: https://codereview.chromium.org/2165523003

Fix GrGLInterface::validate() to check for GL_OES_sample_shading on ES contexts
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2156343004

Review-Url: https://codereview.chromium.org/2156343004

Delete the legacy SVG parser

Unused.

R=reed@google.com,robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2154193003
CQ_INCLUDE_TRYBOTS=master.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/2154193003

GN: polyfill is_fuchsia

I'll tell you what, I need to practice typing fuchsia out a few hundred
times...  I managed to spell it three different ways in this CL.

Plus, gn format BUILD.gn

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2164453003

Review-Url: https://codereview.chromium.org/2164453003

Delete SkXMLPullParser

Unused & not being built.

R=reed@google.com,robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2163463002

Review-Url: https://codereview.chromium.org/2163463002

Rename variables in SkCanvas::internalSaveLayer

This is split out of https://codereview.chromium.org/2145343005/ (Recast draw_filter_into_device as drawDevice) to, hopefully, make it easier to integrate that CL

TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2162883002

Review-Url: https://codereview.chromium.org/2162883002

Fix the assert and clamping in SkMatrix::get*Scale[s]().

Follow-up after https://codereview.chromium.org/2143133005/.

BUG=skia:4718
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2154123003

Review-Url: https://codereview.chromium.org/2154123003

Fix rewinding bug in SkJpegCodec

Performing a sampled and/or subset decode will create some state
in SkJpegCodec.  If we fail to clean up this state properly,
subsequent decodes may try to reuse (and potentailly overflow)
the leftover memory.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2161593003

Committed: https://skia.googlesource.com/skia/+/4ecb8ab556214c9337f56bc36d50e4d7c655ac7a
Review-Url: https://codereview.chromium.org/2161593003

Rotate bitmap strikes with FreeType.

BUG=skia:3490
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2139703002

Committed: https://skia.googlesource.com/skia/+/31e0c1379e6d0ce48196183e295b929af51fa74e
Review-Url: https://codereview.chromium.org/2139703002

Revert of Fix rewinding bug in SkJpegCodec (patchset #1 id:1 of https://codereview.chromium.org/2161593003/ )

Reason for revert:
Breaking Google3 build.  Will reland corrected version.

Original issue's description:
> Fix rewinding bug in SkJpegCodec
>
> Performing a sampled and/or subset decode will create some state
> in SkJpegCodec.  If we fail to clean up this state properly,
> subsequent decodes may try to reuse (and potentailly overflow)
> the leftover memory.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2161593003
>
> Committed: https://skia.googlesource.com/skia/+/4ecb8ab556214c9337f56bc36d50e4d7c655ac7a

TBR=djsollen@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/2155163004

Remove guards in GrGLInterface::validate() for functions that Chromium used to not provide.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2158983002

Review-Url: https://codereview.chromium.org/2158983002

Push GrPipelineBuilder down to the drawBatch calls in the TextContext

This is split out of https://codereview.chromium.org/2092893003/ (Retract PipelineBuilder some more)

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2155983003

Review-Url: https://codereview.chromium.org/2155983003

[SkXMLParser] Initial text node support

Also disable entity processing.

R=bungeman@google.com,robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2154853002

Review-Url: https://codereview.chromium.org/2154853002

Add bounds info

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2159853003

Review-Url: https://codereview.chromium.org/2159853003

check for culled-out paths inside SkDraw

BUG=629026
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2155213002

Review-Url: https://codereview.chromium.org/2155213002

disable test that fails to compute bounds on nexus 5

TBR=jcgregario@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2160713003

Review-Url: https://codereview.chromium.org/2160713003

disable new test since it fails on skia_fast

TBR=jcgregario@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2155243002

Review-Url: https://codereview.chromium.org/2155243002

Tweak a couple things I found confusing in ColorSpaceXform unit tests.

No coverage changes... just refactoring.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2158033002

Review-Url: https://codereview.chromium.org/2158033002

Remove SkDOM::build(const char doc[], size_t len)

This flavor is unused, and we have the more general SkStream API.

R=stephana@google.com,reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2154733003

Review-Url: https://codereview.chromium.org/2154733003

Introduce GrColorSpaceXform, for gamut conversion on textures

GrTextureAccess optionally includes an instance, computed from the src
and dst color spaces. In all common cases (no color space for either src
or dst, or same color space for both), no object is allocated.

This change is orthogonal to my attempts to get color space attached to
render targets - regardless of how we choose to do that, this will give
us the source color space at all points where we are connecting src to
dst.

There are many dangling injection points where I've been inserting
nullptr, but I have a record of all of them. Additionally, there are now
three places (the most common simple paths for bitmap/image rendering)
where things are plumbed enough that I expect to have access to the dst
color space (all marked with XFORMTODO).

In addition to getting the dst color space, I need to inject shader code
and uniform uploading for appendTextureLookup and friends.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2154753003

Review-Url: https://codereview.chromium.org/2154753003

Convert SK Images to new assets format

Depends on https://codereview.chromium.org/2162563002/

BUG=skia:5427
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2155163002

Review-Url: https://codereview.chromium.org/2155163002

Fix rewinding bug in SkJpegCodec

Performing a sampled and/or subset decode will create some state
in SkJpegCodec.  If we fail to clean up this state properly,
subsequent decodes may try to reuse (and potentailly overflow)
the leftover memory.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2161593003

Review-Url: https://codereview.chromium.org/2161593003

Initialize fGammaCorrect in DF Path Renderer unit test

TBR=jvanverth@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2157933003

Review-Url: https://codereview.chromium.org/2157933003

pathops coincidence and security rewrite

Most changes stem from working on an examples bracketed
by #if DEBUG_UNDER_DEVELOPMENT  // tiger
These exposed many problems with coincident curves,
as well as errors throughout the code.

Fixing these errors also fixed a number of fuzzer-inspired
bug reports.

* Line/Curve Intersections
Check to see if the end of the line nearly intersects
the curve. This was a FIXME in the old code.

* Performance
Use a central chunk allocator.
Plumb the allocator into the global variable state
so that it can be shared. (Note that 'SkGlobalState'
is allocated on the stack and is visible to children
functions but not other threads.)

* Refactor
Let SkOpAngle grow up from a structure to a class.
Let SkCoincidentSpans grow up from a structure to a class.
Rename enum Alias to AliasMatch.

* Coincidence Rewrite
Add more debugging to coincidence detection.
Parallel debugging routines have read-only logic to report
the current coincidence state so that steps through the
logic can expose whether things got better or worse.

More functions can error-out and cause the pathops
engine to non-destructively exit.

* Accuracy
Remove code that adjusted point locations. Instead,
offset the curve part so that sorted curves all use
the same origin.
Reduce the size (and influence) of magic numbers.

* Testing
The debug suite with verify and the full release suite
./out/Debug/pathops_unittest -v -V
./out/Release/pathops_unittest -v -V -x
expose one error. That error is captured as cubics_d3.
This error exists in the checked in code as well.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003

Review-Url: https://codereview.chromium.org/2128633003

Add makeSpecial calls to SkGpuDevice
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2161533003

Review-Url: https://codereview.chromium.org/2161533003

Add filename to gradient scrape

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2160583002

Review-Url: https://codereview.chromium.org/2160583002

Revert of remove unused TableMaskFilter (patchset #1 id:1 of https://codereview.chromium.org/2156463002/ )

Reason for revert:
Needed for Android:

frameworks/base/core/jni/android/graphics/MaskFilter.cpp:5:10: fatal error: 'SkTableMaskFilter.h' file not found

Original issue's description:
> remove unused TableMaskFilter
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2156463002
>
> TBR=
>
> Committed: https://skia.googlesource.com/skia/+/59779ae8ce316bf8b8082ec2df1683ccd38161f1

TBR=reed@google.com
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG=skia:

Review-Url: https://codereview.chromium.org/2156183002

In GrShape detect that stroked axis-aligned lines are rrects.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2151313002

Review-Url: https://codereview.chromium.org/2151313002

Test auto merger

TBR=djsollen@google.com
NOTRY=true
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2158793004

Review-Url: https://codereview.chromium.org/2158793004

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=
NO_MERGE_BUILDS
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2157823002

Review-Url: https://codereview.chromium.org/2157823002

Guard SkMatrix::get*Scale[s]() against negative nearly-zero values.

BUG=skia:4718
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2143133005

Review-Url: https://codereview.chromium.org/2143133005

Revert of Rotate emoji with FreeType. (patchset #5 id:80001 of https://codereview.chromium.org/2139703002/ )

Reason for revert:
Causing roll to fail on telemetry_perf_unittests (bencharks.system_health_smoke_test.SystemHealthBenchmarkSmokeTest.system_health.memory_desktop.load:search:taobao (and baidu)) and browser_tests (FindInPageControllerTest.FindInPageSpecialURLS).

This is due to triggering the assert in copyFTBitmap

SkASSERT(dstMask.fBounds.width() == static_cast<int>(srcFTBitmap.width));

when called from inside the block guarded by

if (bitmapTransform.isIdentity())

Original issue's description:
> Rotate bitmap strikes with FreeType.
>
> BUG=skia:3490
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2139703002
>
> Committed: https://skia.googlesource.com/skia/+/31e0c1379e6d0ce48196183e295b929af51fa74e

TBR=mtklein@google.com,reed@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:3490

Review-Url: https://codereview.chromium.org/2149253005

Add support for Fuchsia
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2152273002

[mtklein edit from here down]
No public API changes.
TBR=reed@google.com

Review-Url: https://codereview.chromium.org/2152273002

Rotate bitmap strikes with FreeType.

BUG=skia:3490
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2139703002

Review-Url: https://codereview.chromium.org/2139703002

move responsibility for creating raster-device into pdf

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2146333004

NOTRY=True
win failure seems unrelated

Review-Url: https://codereview.chromium.org/2146333004

Speculatively fix Android automerger

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2155713004

Review-Url: https://codereview.chromium.org/2155713004

Add capability for SkColorXform to output half floats

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2147763002
CQ_INCLUDE_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review-Url: https://codereview.chromium.org/2147763002

Dont allocate a new image which already has mip levels, when regenerating vulkan mipmaps

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2146063008

Review-Url: https://codereview.chromium.org/2146063008

SkPdf: smaller color serialization

SkPDFUtils now has a special function (SkPDFUtils::AppendColorComponent)
just for writing out (color/255) as a decimal with three digits of
precision.

SkPDFUnion now has a type to represent a color component.  It holds a
utint_8, but calls into AppendColorComponent to serialize.

Added a unit test that tests all possible input values.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2151863003

Review-Url: https://codereview.chromium.org/2151863003

SkPDF: Join Positioned Text

When N sequential positioned glyphs differ in positions by exactly the
advances of the first (N-1) glyphs, join the glyphs into a string
rather than changing the text matrix between each glyph draw.

Decreases PDF output size by about ~1.4%.  Potentially more on
text-heavy pages.

A single-typeface PDF of an 27kB ASCII document shaped with harfbuzz:
        before:  187743 Bytes
         after:  65513 Bytes
    difference:  -65.1%

Before:
    BT
    /F0 13 Tf
    1 0 0 -1 143.5 61 Tm
    <0029> Tj
    1 0 0 -1 150.634765 61 Tm
    <004C> Tj
    1 0 0 -1 154.602050 61 Tm
    <0055> Tj
    1 0 0 -1 160.245117 61 Tm
    <0048> Tj
    1 0 0 -1 167.925781 61 Tm
    <004B> Tj
    1 0 0 -1 176.469726 61 Tm
    <0052> Tj
    1 0 0 -1 184.518554 61 Tm
    <0056> Tj
    1 0 0 -1 190.980468 61 Tm
    <0048> Tj
    ET

After:
    BT
    /F0 13 Tf
    1 0 0 -1 0 0 Tm
    143.5 -61 Td <0029004C0055> Tj
    16.7451171 0 Td <0048004B005200560048> Tj
    ET

Also: update the Text matrix with the `Td` operator, instead of
overwriting it with the the `Tm` operator.  In the worst case, when
every glyph is positioned differently than it's advance, this still
makes the command stream smaller:

Before:
    ...
    1 0 0 -1 58.328125 660 Tm <0055> Tj
    1 0 0 -1 61.609375 660 Tm <004C> Tj
    1 0 0 -1 63.828125 660 Tm <0056> Tj
    ...

After:
    ...
    3.140625 0 Td <0055> Tj
    3.28125 0 Td <004C> Tj
    2.21875 0 Td <0056> Tj
    ...

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2150393002

Review-Url: https://codereview.chromium.org/2150393002

SkPDF: Always give a default advance for glyphs.

Prior to this we assumed the default default was 0.  It is
actually 1000, according to the spec.

fixes BUG=skia:5321
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2146403005

Review-Url: https://codereview.chromium.org/2146403005

Fix broken Android framework build.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2153953002

Review-Url: https://codereview.chromium.org/2153953002

Add copy (when needed) to SkGpuDevice::asSpecial

split out of https://codereview.chromium.org/2145343005/ (Recast draw_filter_into_device as drawDevice)

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2157513002

Review-Url: https://codereview.chromium.org/2157513002

Exclude XML and dependencies from Bazel build.

Depends on https://codereview.chromium.org/2150353002

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2147303004

Review-Url: https://codereview.chromium.org/2147303004

SkColorSpace::NewICC() warn instead of fail on bad input

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2093193002

Review-Url: https://codereview.chromium.org/2093193002

Make XML compilation conditional.

Only enabled for gyp-based builds at this time.

R=dogben@google.com,mtklein@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2150353002

Review-Url: https://codereview.chromium.org/2150353002

remove unused CreateInfo flag

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2153173002

NOTRY=True

Review-Url: https://codereview.chromium.org/2153173002

Convert Win toolchain to CIPD package

BUG=skia:5427
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2111713003

Review-Url: https://codereview.chromium.org/2111713003

Add SkGpuDevice::asSpecial entry point
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2151033003

Review-Url: https://codereview.chromium.org/2151033003

Add a bench to measure the best way to pack from int to uint16_t with SSE.

I measured relative runtimes on my laptop:

   pack_int_uint16_t_ss…
   1036  …e41 1x  …se3 1.01x  …e2_b 3.01x  …e2_a 3.02x

I've run into Clang problems with the actual _mm_packus_epi32 instruction, I think,
so I'm going to exercise a little cowardice and leave that option disabled for now.

The ssse3 version probably looks a little faster than it will be in practice.
We'll usually need to load its mask, which here is hoisted out of the bench loop.

The two sse2 variants are close enough in speed that I'm tie breaking them on other
concerns: the <<16, >>16 version doesn't need any scratch registers or to load any
constants, so it wins.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2150343002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot,Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast-Trybot

Review-Url: https://codereview.chromium.org/2150343002

Expand _01 half<->float limitation to _finite.  Simplify.

    It's become clear we need to sometimes deal with values <0 or >1.
    I'm not yet convinced we care about NaN or +-inf.

    We had some fairly clever tricks and optimizations here for NEON
    and SSE.  I've thrown them out in favor of a single implementation.
    If we find the specializations mattered, we can certainly figure out
    how to extend them to this new range/domain.

    This happens to add a vectorized float -> half for ARMv7, which was
    missing from the _01 version.  (The SSE strategy was not portable to
    platforms that flush denorm floats to zero.)

    I've tested the full float range for FloatToHalf on my desktop and a 5x.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2145663003
CQ_INCLUDE_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot,Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast-Trybot

Committed: https://skia.googlesource.com/skia/+/3296bee70d074bb8094b3229dbe12fa016657e90
Review-Url: https://codereview.chromium.org/2145663003

Add getColorSpace to GrTextureProducer

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2150113002

Review-Url: https://codereview.chromium.org/2150113002

Probably no need for GM self-tests now that GM doesn't exist.

This is... belated.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2154603002

Review-Url: https://codereview.chromium.org/2154603002

Dust-off SkXMLParser

Hook up SkXMLParser to Expat, such that it can actually parse, err,
XML.

Add a trivial unit test.

R=robertphillips@google.com,reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2142893006

Review-Url: https://codereview.chromium.org/2142893006

remove unused TableMaskFilter

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2156463002

TBR=

Review-Url: https://codereview.chromium.org/2156463002

Fix a bug in the Clamp in X direction tiling.
The code mixed up which end of the span was cut and preserved in the
sequence of span breaks.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2145193002

Review-Url: https://codereview.chromium.org/2145193002

Remove GrWrapTextureInBitmap call in SkGpuDevice::drawSpriteWithFilter

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2146223003

Review-Url: https://codereview.chromium.org/2146223003

Build Expat on all platforms

* bump version to @android-6.0.1_r55 (2.1.1)

* add as a dependency to the 'xml' target (for upcoming http://crrev.com/2142893006)

* tweak build for Mac, Win:

  - define HAVE_EXPAT_CONFIG_H (same as in Android.mk)
  - use -Wno-missing-field-initializers (same as in Android.mk)
  - suppress MSVS warning 4244 (same as ..cmake/Utilities/cmexpat/expatConfig.h.in)

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2150603004
R=mtklein@google.com

Review-Url: https://codereview.chromium.org/2150603004

Remove dead flag from SkPictureRecorder

With the Chrome CL (https://codereview.chromium.org/2149533003/ (Remove use of unsupported Skia flag)) landed, this should be good to go.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2152943002

Review-Url: https://codereview.chromium.org/2152943002

Revert of Expand _01 half<->float limitation to _finite.  Simplify. (patchset #7 id:120001 of https://codereview.chromium.org/2145663003/ )

Reason for revert:
Unit tests fail on Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast

Original issue's description:
> Expand _01 half<->float limitation to _finite.  Simplify.
>
>     It's become clear we need to sometimes deal with values <0 or >1.
>     I'm not yet convinced we care about NaN or +-inf.
>
>     We had some fairly clever tricks and optimizations here for NEON
>     and SSE.  I've thrown them out in favor of a single implementation.
>     If we find the specializations mattered, we can certainly figure out
>     how to extend them to this new range/domain.
>
>     This happens to add a vectorized float -> half for ARMv7, which was
>     missing from the _01 version.  (The SSE strategy was not portable to
>     platforms that flush denorm floats to zero.)
>
>     I've tested the full float range for FloatToHalf on my desktop and a 5x.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2145663003
> CQ_INCLUDE_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot;master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot,Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast-Trybot
>
> Committed: https://skia.googlesource.com/skia/+/3296bee70d074bb8094b3229dbe12fa016657e90

TBR=msarett@google.com,mtklein@chromium.org
# 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/2151023003