path ops stream-lining

The addT() function is a workhorse of pathops.
Make it simpler, removing branches and parameters.

Separate addOpp() into const and modify parts.

Add more debugging that asserts if the function
fails and the data is not extreme (e.g., fuzzer
generated).

TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2273293004

Review-Url: https://codereview.chromium.org/2273293004

SkDrawCommand: hinting

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2274373003

Review-Url: https://codereview.chromium.org/2274373003

BUILD.gn: Fix fiddle raster.

BUG=skia:

Review-Url: https://codereview.chromium.org/2269143007

SkPDF: Stop `#include PREPROCESSOR_DEFINE` pattern

No more:
   #include SK_SFNTLY_SUBSETTER
   #include ZLIB_INCLUDE

Also, rename SK_SFNTLY_SUBSETTER to SK_PDF_USE_SFNTLY
to follow my pattern of prefixing SkPDF-specific defines
with 'SK_PDF_'.

The ZLIB_INCLUDE define is no longer is used by anyone.

TODO: rename Sfntly to something pronounceable.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2273343002

Review-Url: https://codereview.chromium.org/2273343002

Remove pixel config fallback - failing is a better option.

This produces surprising results in some cases. For example: Attempt to
make an F16 SkSurface on ANGLE. We can't render to F16, but this succeeds
and gives you back an N32 surface instead. Ran all tests and GMs in DM on
ANGLE (where this is most likely to be an issue). No problems.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2272423004

Review-Url: https://codereview.chromium.org/2272423004

update Android auto-detection.

Clang defines __ANDROID__

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2278483004

No public API changes.
TBR=reed@google.com

Review-Url: https://codereview.chromium.org/2278483004

Reland: Experimental parsing expression grammar (PEG) template library

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2271743002

Committed: https://skia.googlesource.com/skia/+/9d08cbc8c6131ff61a1e71cc5c8cf27841d62b42
Review-Url: https://codereview.chromium.org/2271743002

Add Ganesh support for circular roundrects with strokes > 2*radii.

Overstroked rrects are supported by adding additional geometry to the
center of the nine patch, and setting constant offset vectors to fake
out the distance calculation.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2272703004

Review-Url: https://codereview.chromium.org/2272703004

Make bleed GM produce consistent bitmaps on all platforms
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2264133003

Review-Url: https://codereview.chromium.org/2264133003

Made shadows blurry (thru implementing variance mapping)

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2224163005

Review-Url: https://codereview.chromium.org/2224163005

Update Skia milestone to 55

TBR=reed@google.com

No API changes, just moving version fwd.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2281533002

Review-Url: https://codereview.chromium.org/2281533002

Add fat stroke test case to roundrects GM.

Prep for supporting fat stroked roundrects in Ganesh.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2272703009

Review-Url: https://codereview.chromium.org/2272703009

Revert of Experimental parsing expression grammar (PEG) template library (patchset #8 id:140001 of https://codereview.chromium.org/2271743002/ )

Reason for revert:
G3 roll & Msan woes.

Original issue's description:
> Experimental parsing expression grammar (PEG) template library
>
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2271743002
>
> Committed: https://skia.googlesource.com/skia/+/9d08cbc8c6131ff61a1e71cc5c8cf27841d62b42

TBR=mtklein@google.com,bungeman@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:

Review-Url: https://codereview.chromium.org/2275943004

add pathops debugging

Pathops has many points of failure, most of which
are triggered by extreme data generated by fuzzers.
It's difficult to figure out which failure point
was triggered when the operation gives up.

Add instrumentation so that the failure can
be debugged when the data is well-behaved.

Also, add a check that looks for a sequence of
coincident points on multiple edges that are out
of order when compared to each other.

TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2274803003

Review-Url: https://codereview.chromium.org/2274803003

Add neon and crc32 sources for aarch64

This fixes the build for aarch64 arch

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2272243003

Review-Url: https://codereview.chromium.org/2272243003

Experimental parsing expression grammar (PEG) template library

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2271743002

Review-Url: https://codereview.chromium.org/2271743002

Update GrShape test to allow more flexible shape creation.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2277483002

Review-Url: https://codereview.chromium.org/2277483002

GN: misc

  - Use options' template pattern for opts too.
  - Simplify opt's and options' configs... they should all be the same.
  - When building a static-library component in our GN environment (i.e. libskia.a),
    make it a complete static lib, fully containing its transitive deps.
  - It has not proved useful to override ar.

TBR=jcgregorio@google.com

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2278673002

Review-Url: https://codereview.chromium.org/2278673002

We can't infer the right type for voidp for old png_jmpbuf()

In libpng 1.2 it's just a macro that's ->jmpbuf,
so there's nothing forcing the conversion to png_structp.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2268953005

Review-Url: https://codereview.chromium.org/2268953005

SkPDF: set SK_SFNTLY_SUBSETTER uniformly

When Skia is built into the Android framework,
SK_SFNTLY_SUBSETTER="sample/chromium/font_subsetter.h".
This #includes the same value for GOOGLE3.

Once Chrome is also using this value, we can do away with
the `#ifdef GOOGLE3` line.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2272103002

Review-Url: https://codereview.chromium.org/2272103002

mark fuzz test flaky since it may fail on some platforms

NOTRY=true
TBR=bungeman@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2277893004

Review-Url: https://codereview.chromium.org/2277893004

gyp/sfntly: set SK_SFNTLY_SUBSETTER uniformly

When Skia is built into the Android framework,
SK_SFNTLY_SUBSETTER="sample/chromium/font_subsetter.h".
This sets the same value for Skia's test framework

I will eventually move away from the
   #include PREPROCESSOR_DEFINE
trick, which does not work everywhere.

No public API changes.
TBR=reed@google.com

Review-Url: https://codereview.chromium.org/2274223002

GN: guard tools (except fiddle) by skia_enable_tools.

Our tools use third-party libraries pretty freely, some of which may not
be available in other GN environments (e.g. Fuchsia).  Most can also not
function when Skia is built as a shared library.

fiddle stands alone as the exception to both those points: it depends on
only Skia, and works fine with both a shared or static library.

So guard everything but fiddle with this flag skia_enable_tools, disabled
when we're building for Fuchsia or when we're build a shared library.

This CL has a couple of little tweaks to Fiddle to keep it working:
  - divorce it from :tool_utils, instead just building SkForceLinking.cpp itself;
  - fix up a buggy rebase_path() that was accidentally working when we depended
    on :tool_utils;
  - drop test_only: it now only requires production-code dependencies.

The SkImageEncoder Create* methods need to be SK_API if we want SkForceLinking
to work across .so's.  Without this, SkForceLinking needs to be part of Skia; it
can't be part of the application using Skia.

The rest is mostly just a re-indent under if (skia_enable_tools),
courtesy of gn format.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2273823003

No public API changes.
TBR=reed@google.com

Review-Url: https://codereview.chromium.org/2273823003

SkPngCodec: voidp instead of forward-declares for png.h types.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2277903002

Review-Url: https://codereview.chromium.org/2277903002

Fix leak in GrTextureStripAtlasTest

This is a followup to https://codereview.chromium.org/2262233002/,
adding a missing call to GrTextureStripAtlas::unlockRow.

BUG=chromium:637678
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2272953003

Review-Url: https://codereview.chromium.org/2272953003

s/invertable/invertible

English orthography is often made difficult by poor phonology.

TBR=oxford-english-dictionary,mtklein

Review-Url: https://codereview.chromium.org/2272083002

tiny line breaks gl render

This pie cut draws correctly
on raster but extends to the
right on Ganesh.

The path is not convex, and
has a tiny line that contributes
to the error.

TBR=robertphillips@google.com
BUG=640176
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2276083002

Review-Url: https://codereview.chromium.org/2276083002

Fix SkTLazy(const T*) initialization

Splitting the fix from https://codereview.chromium.org/2271743002/.

R=bungeman@google.com,mtklein@google.com
TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2278633002

Review-Url: https://codereview.chromium.org/2278633002

GN: Fuchsia probably cannot link without this.

We removed this _none variant globally, thinking we'd put back platform
specific versions as we added support for those platforms.

We don't have anything for Fuchsia.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2271273002

Review-Url: https://codereview.chromium.org/2271273002

SkPDF: vector canvases can't hint!

BUG=637571
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2273073002

Review-Url: https://codereview.chromium.org/2273073002

GN: more optional components: jpeg, pdf, png, xml

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2267343004

Review-Url: https://codereview.chromium.org/2267343004

remove point aliases

This removes the notion of keeping track of every different t value
that resolves to the same or a similar point. Other fixes make
this concept unnecessary, and removing it simplifies the code.

This removes an allocation, and speeds up paths with many
overlapping curves.

As a bonus, four fuzzer tests that failed before now succeed.

TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2275703003

Review-Url: https://codereview.chromium.org/2275703003

SkPDF: vertical writing: draw nothing

BUG=skia:5665
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2274153002

Review-Url: https://codereview.chromium.org/2274153002

Add simple font fallback on Mac.

Implement SkFontMgr_Mac::onMatchFamilyStyleCharacter. This is a simple
implementation which provides basic fallback.

This also renames and refactors several static functions to reduce code
duplication.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2268663002

Review-Url: https://codereview.chromium.org/2268663002

Make GrTextureStripAtlas flush pending IO on newly acquired texture

GrTextureStripAtlas uses its own lock counts to protect against
overwriting its own earlier writes, but that doesn't protect against
IO that was pending when a texture was first acquired.

BUG=chromium:637678
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2262233002

Review-Url: https://codereview.chromium.org/2262233002

Parse ICC profiles from webps

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2269333002

Review-Url: https://codereview.chromium.org/2269333002

Add addtional resolve calls to vulkan backend

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2215363003

Review-Url: https://codereview.chromium.org/2215363003

Warn on bad rendering intents (instead of valid ones)

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2274843002

Review-Url: https://codereview.chromium.org/2274843002

New helper for creating a color space from another, with linear gamma

Going to need this in many places once we expliclty require F16 surfaces
to have linear gamma color spaces.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2270103002

Review-Url: https://codereview.chromium.org/2270103002

Apply gerrit ref if it is a Gerrit patch.

Also:
* Pass through a new property 'patch_storage' to DM/Nanobench/Coverage. This will be used by the different frameworks to figure out if it is Rietveld or Gerrit issue.
* Calculate issue and patchset for Gerrit patches similar to Rietveld.

BUG=skia:5627
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2263323002

Review-Url: https://codereview.chromium.org/2263323002

combine setRectFan and mapRect

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2276603002

Review-Url: https://codereview.chromium.org/2276603002

Fix generic color space xform, ColorSpaceXformTest

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2275563002

Review-Url: https://codereview.chromium.org/2275563002

Make singleton SkColorSpaces thread safe

SkColorSpaces may be once-ptrs, so we can reuse common color spaces.
This means that they must be thread safe.  SkMatrix44 is not
thread safe because it maintains a mutable type mask.

This CL ensures that we precompute the type mask so we
can use const SkMatrix44's safely.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2277463002

Review-Url: https://codereview.chromium.org/2277463002

Remove SkPixelRef::getTexture()
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2264293003

Review-Url: https://codereview.chromium.org/2264293003

GN: extract optional() as a template and use for giflib-dependent code

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2270833003

Review-Url: https://codereview.chromium.org/2270833003

GN: Fix up Ganesh native interface hooks.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2269213002

Review-Url: https://codereview.chromium.org/2269213002

Skip non-AA intersect rects in GrReducedClip

Skips non-AA rects whose op is intersect or replace, and who do not
precede elements that grow the clip, by tightening fIBounds.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2271493002

Review-Url: https://codereview.chromium.org/2271493002

SkScalerContextRec::computeMatrices to return status.

When computing the matricies for a scaler context, there is a special
case when the matrix is determined to be singular. No port properly
handles zero sized text, so we detect this case and return a 'normal'
text size and a zero matrix for all computed transformations. This
CL causes computeMatricies to return 'false' in this case.

This is used in the constructor of SkScalerContext_Mac in order to
avoid calling CGAffineTransformInvert on non-invertible transformations.
CGAffineTransformInvert documents that if the transform is
non-invertible it will return the passed transform unchanged. It does
so, but then also prints a message to stdout. Since the information is
already available to avoid this chatty behavior, use it to keep things
quiet.

BUG=skia:3231,chromium:630169
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2276583003

Review-Url: https://codereview.chromium.org/2276583003

Assume all TURBO_HAS_* are true.

Should be everyone's on libjpeg-turbo >= 1.5.0.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2274643002

Review-Url: https://codereview.chromium.org/2274643002

Moving SkBlurImageFilter into core

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003

Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48
Committed: https://skia.googlesource.com/skia/+/e426babe7552b1cb4e27cdf4e90826feabb9e3b0
Committed: https://skia.googlesource.com/skia/+/4d760175686df8f61a11b66946eb307d89dd2c75
Committed: https://skia.googlesource.com/skia/+/a97a1ab5719d5c355f7900b7f17dec1e467cf57e
Review-Url: https://codereview.chromium.org/2255803003

CMake: require libjpeg-turbo, not libjpeg

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2273753002
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/2273753002

experimental/tools/mskp_parser.py

TBR=
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2268403003

Review-Url: https://codereview.chromium.org/2268403003

Revert of Moving SkBlurImageFilter into core (patchset #9 id:160001 of https://codereview.chromium.org/2255803003/ )

Reason for revert:
Breaking internal roll. Still needs 'Create'.

Original issue's description:
> Moving SkBlurImageFilter into core
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003
>
> Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48
> Committed: https://skia.googlesource.com/skia/+/e426babe7552b1cb4e27cdf4e90826feabb9e3b0
> Committed: https://skia.googlesource.com/skia/+/4d760175686df8f61a11b66946eb307d89dd2c75
> Committed: https://skia.googlesource.com/skia/+/a97a1ab5719d5c355f7900b7f17dec1e467cf57e

TBR=reed@google.com,fmalita@chromium.org,vjiaoblack@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/2274603003

fix conic path fuzz

The test conic has a very large weight, so it reduces to a pair
of lines. Detect this case rather than subdividing the conic
so much that the answers are meaningless.

R=herb@google.com, reed@google.com
BUG=638223
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2268073004

Review-Url: https://codereview.chromium.org/2268073004

SkMultiSKP: version 2

Measurable size improvement.

BUG=skia:5370

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255333003

Review-Url: https://codereview.chromium.org/2255333003

SkPDF: PDFDevice cleanup.

Remove unused fContentSize.

Combine SkPDFDevice::writeContent and SkPDFDevice::content.

Remove unused SkPDFDevice::initialTransform().

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2264113003

Review-Url: https://codereview.chromium.org/2264113003

f16<->f32 ftz is an optional thing for speed.

The ARMv8 asm path actually does it right... that should be okay.

My Nexus 5x fails `dm -m _finite_ftz` before this and passes after it.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2276533002

TBR=msarett@google.com

Review-Url: https://codereview.chromium.org/2276533002

Moving SkBlurImageFilter into core

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003

Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48
Committed: https://skia.googlesource.com/skia/+/e426babe7552b1cb4e27cdf4e90826feabb9e3b0
Committed: https://skia.googlesource.com/skia/+/4d760175686df8f61a11b66946eb307d89dd2c75
Review-Url: https://codereview.chromium.org/2255803003

fix pathops fuzz

Extreme cubics may split so that one half is a point.
Discard this rather than generating a degenerate line.

TBR=reed@google.com
BUG=640025
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2276503002

Review-Url: https://codereview.chromium.org/2276503002

GN: make libwebp an optional dependency

This will be handy for folks who don't have libwebp, like Fuchsia.

I convinced myself that this is done right by:
  - building in all three modes (default and explicitly set both ways);
  - looking at verbose Ninja logs to see the presence/lack of SK_HAS_WEBP_LIBRARY;
  - running dm -m Codec, which passes with libwebp and segfault without it.

If this is viable, I intend to make all third-party dependencies optional
and follow this pattern.  :skia should link and degrade gracefully without
any of //third_party.  It's okay for tools to have hard third-party
dependencies; we just need them to get past the `gn gen` stage without them.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2270693004

Review-Url: https://codereview.chromium.org/2270693004

Reduce window rectangles cap to 8

Lowers the cap to 8 and adds a warning message if this value is ever
exceeded. The largest (only) implementation currently supports 8, so
there isn't yet reason to go higher.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2254013002

Committed: https://skia.googlesource.com/skia/+/52d721580ee22525c285e2d13cf3975a7a1b2843
Review-Url: https://codereview.chromium.org/2254013002

Cleanup use of legacy SkSurface creation methods a bit

This CL cleans up some remaining uses of the old API

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2223023002

Committed: https://skia.googlesource.com/skia/+/f21cd16228c20927fa4e9b937d6951471cfaa37a
Review-Url: https://codereview.chromium.org/2223023002

Fix valgrind issue in CircleBatch__Test

TBR=bsalomon@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2271653002

Review-Url: https://codereview.chromium.org/2271653002

Remove file samplecode/SampleDraw.cpp

The file is not used in any build target and seems dead, i.e. failed to compile when I added it to
the SampleApp target.

BUG=None
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2269583002

Review-Url: https://codereview.chromium.org/2269583002

Fix Ganesh analytic blurred rect draws

This CL does two things:
 It fixes the SkBlurMaskFilterImpl::directFilterRRectMaskGPU draw path to explicitly handle rects
 It fixes the SkGpuDevice::drawTextureProducerImpl draw path to provide the correct (src & device space) inputs to directFilterRRectMaskGPU.

How this was working before was that GrRRectBlurEffect::Make would reject rect-rrects and the code would fallback to
GrBlurUtils::drawPathWithMaskFilter which mapped the rect-rrect into device space correctly (of course,
the rect-ness of the path was removed at that point so it was going through the slow path).

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2268583002

Review-Url: https://codereview.chromium.org/2268583002

Fix GLSL int to float error in arc code in GrOvalRenderer.

BUG=skia:
TBR=robertphillips@google.com
NOTRY=true
NOTREECHECKS=true

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2263213004

Review-Url: https://codereview.chromium.org/2263213004

More robust check for sRGB gamma tables

This is in response to a UMA showing that 5% dst gammas are
unidentified tables.  We want to see if some of these tables
should be marked as sRGB.
https://uma.googleplex.com/p/chrome/histograms?endDate=latest&dayCount=1&histograms=Blink.ColorSpace.Destination&fixupData=true&showMax=true&filters=isofficial%2Ceq%2CTrue&implicitFilters=isofficial

This check is not fast.  If we find that it doesn't help
us recognize sRGB curves, we should delete it.

BUG=skia:5656
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2263233003

Review-Url: https://codereview.chromium.org/2263233003

Remove virtual from overrides in SkGpuDevice.h
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2265323002

Review-Url: https://codereview.chromium.org/2265323002

Fix CircleBatch::TestCreate to always use a legal matrix

TBR=robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2267953002

NOTREECHECKS=true
NOTRY=true

Review-Url: https://codereview.chromium.org/2267953002

Mark temporary paths created by SkCanvas for clipping as volatile
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2268723002

Review-Url: https://codereview.chromium.org/2268723002

Remove vulkan from most build variants

Since many devices don't support vulkan, we have to remove it for most
build variants and only turn it on manually by selecting arm64vulkan
build variant.

BUG=skia:5516
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255973003

Review-Url: https://codereview.chromium.org/2255973003

Flush denorm half floats to zero.

I think we convinced ourselves that denorms, while a good chunk of half floats,
cover a rather small fraction of the representable range, which is always
close enough to zero to flush.

This makes both paths of the conversion to or from float considerably simpler.

These functions now work for zero-or-normal half floats (excluding infinite, NaN).
I'm not aware of a term for this class so I've called them "ordinary".

A handful of GMs and SKPs draw differently in --config f16, but all imperceptibly.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2256023002

Review-Url: https://codereview.chromium.org/2256023002

GPU implementation of drawArc.

This adds analytic shaders for filled arcs and butt-cap stroked arcs where the center point is not used.

BUG=skia:5227
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2259323003

Review-Url: https://codereview.chromium.org/2259323003

Add cap type variations to circular arcs GMs

BUG=skia:5227
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2269723002

Review-Url: https://codereview.chromium.org/2269723002

check for wide-open inside quickContains

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2263373002

Review-Url: https://codereview.chromium.org/2263373002

Fix MSAN - mark temporary memory as kNo_ZeroInitialized

TBR=mtklein@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2261173004

Review-Url: https://codereview.chromium.org/2261173004

Add trooper documentation for CT bots and for iOS "the service is invalid"

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2263103003
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=2263103003

Review-Url: https://codereview.chromium.org/2263103003

Fix Equals and serialization for rare pngs

PNGs may contain a gAMA chunk that specifies gamma
values as floats.  If so, we will use these floats
to create an SkColorSpace.

This CL fixes Equals(), serialize(), and
Deserialize() to correctly handle SkColorSpaces
with strange gammas, where we are unable to fall
back on the profile data.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2221983002

Review-Url: https://codereview.chromium.org/2221983002

Revert of Moving SkBlurImageFilter into core (patchset #8 id:140001 of https://codereview.chromium.org/2255803003/ )

Reason for revert:
Misnamed function.

Original issue's description:
> Moving SkBlurImageFilter into core
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003
>
> Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48
> Committed: https://skia.googlesource.com/skia/+/e426babe7552b1cb4e27cdf4e90826feabb9e3b0
> Committed: https://skia.googlesource.com/skia/+/4d760175686df8f61a11b66946eb307d89dd2c75

TBR=reed@google.com,fmalita@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/2265263002

Moving SkBlurImageFilter into core

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003

Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48
Committed: https://skia.googlesource.com/skia/+/e426babe7552b1cb4e27cdf4e90826feabb9e3b0
Review-Url: https://codereview.chromium.org/2255803003

Remove quantization of device-space dest rect for Ganesh analytic rrect blurs

At least on Windows, this fixes a lot of the horrible banded blurs. AFAICT there is no good reason to quantize the device space rect passed to the GrRectBlurEffect.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2262263002

Review-Url: https://codereview.chromium.org/2262263002

Mark temp paths in SkBaseDevice::drawArc and ::drawDRRect as volatile
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2261123004

Review-Url: https://codereview.chromium.org/2261123004

Fix touch event parameter order

The issue is so obvious... However, all newer Android devices (e.g.,
Nexus 6P, 5X, 9) are immune to this bug. So I didn't notice this until I
tested Viewer app on Nexus 7 yesterday...

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2254223002

Review-Url: https://codereview.chromium.org/2254223002

Revert of Moving SkBlurImageFilter into core (patchset #8 id:140001 of https://codereview.chromium.org/2255803003/ )

Reason for revert:
Fixed it wrong, needs to revert to re-discuss and re-land.

Original issue's description:
> Moving SkBlurImageFilter into core
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003
>
> Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48
> Committed: https://skia.googlesource.com/skia/+/e426babe7552b1cb4e27cdf4e90826feabb9e3b0

TBR=reed@google.com,fmalita@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/2266063002

Moving SkBlurImageFilter into core

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003

Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48
Review-Url: https://codereview.chromium.org/2255803003

Add stroke-and-fill circular arc test, test near 180deg arc in all circular_arc_tests

BUG=skia:5227
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2261233002

Review-Url: https://codereview.chromium.org/2261233002

Detect all named gammas

Our DstColorSpace UMA is showing some named gammas that are
not appropriately detected and placed in named categories.
https://uma.googleplex.com/p/chrome/histograms?endDate=latest&dayCount=1&histograms=Blink.ColorSpace.Destination&fixupData=true&showMax=true&filters=isofficial%2Ceq%2CTrue&implicitFilters=isofficial

This CL should fix that.

I'm not sure (yet) how I feel about this landing permanently.
Seems a little messy.

But it will be interesting to see how this affects the UMA.
My best guess is that we are hitting this case when all
three gammas are "invalid" in different ways.  I'm expecting
to see some profiles end up in the "invalid" category now.

It's also possible that we'll see these cases being absorbed
into sRGB or somewhere else.

BUG=skia:5656
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2261213002

Review-Url: https://codereview.chromium.org/2261213002

compress_r11eac_blocks() required more alignment than dst has.

This shouldn't change any behavior except that the stores to dst
will no longer require 8-byte alignment.

Empirically it seems like we can use 4-byte alignment here,
but u8 (i.e. 1-byte alignment) is always safe.

BUG=skia:5637
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2264103002
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/2264103002

Revert of Moving SkBlurImageFilter into core (patchset #6 id:100001 of https://codereview.chromium.org/2255803003/ )

Reason for revert:
It looks like this breaks our roll into Google3:
https://test.corp.google.com/ui#id=OCL:130943857:BASE:130944046:1471881622765:61dbdd3

Original issue's description:
> Moving SkBlurImageFilter into core
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003
>
> Committed: https://skia.googlesource.com/skia/+/dd3259eb95c3b47e11eefa3b176365a112a32b48

TBR=reed@google.com,fmalita@chromium.org,vjiaoblack@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/2263283002

Use ARMv8 CRC32 instructions for SkOpts::hash().

For large inputs, this runs ~11x faster than Murmur3.
My bench drops from 1µs to 88ns.

Like x86-64, this runs fastest if we work in 24 byte chunks.  16 byte chunks
run at about 0.75x this speed, 8 byte chunks at about 0.4x (which would still
be about 5x faster than Murmur3).

This'll require plumbing support for opts_crc32 into Chrome first before it can roll.

perf.skia.org charts we want to watch: https://perf.skia.org/#5490

Seach for compute_hash in these logs to see the difference:
baseline: https://luci-milo.appspot.com/swarming/task/30ba22f3dfe30e10/steps/nanobench/0/stdout
trybot: https://luci-milo.appspot.com/swarming/task/30bbc406cbf62d10/steps/nanobench/0/stdout

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2260823002
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/2260823002

Moving SkBlurImageFilter into core

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2255803003

Review-Url: https://codereview.chromium.org/2255803003

Support color xforms for kIndex8 pngs

This change started as: "Always use color xforms to
premultiply".  We need to be in a linear space to
premultiply correctly.

It became clear that we also need to support kIndex8
color xforms in order to make this change.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2246143002

Review-Url: https://codereview.chromium.org/2246143002

'g' key toggles showgrid for PerlinPatch

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2269673002

TBR=

Review-Url: https://codereview.chromium.org/2269673002

Supply a color space in sRGB/F16 mode of SampleApp

Ganesh now uses this to decide if it should perform gamma-correct
rendering, so the color space is necessary to get correct results in
S32/F16.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2266763004

Review-Url: https://codereview.chromium.org/2266763004

Roll recipe DEPS

BUG=skia:5627
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2266933002

Review-Url: https://codereview.chromium.org/2266933002

'F' will toggle filtering for the --picture sample

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2263203002

TBR=

Review-Url: https://codereview.chromium.org/2263203002

Fix color xform width bug when scaling/subsetting

This was not caught by the bots because we don't test
color correct modes with our many image decoding tests
(takes too long).

Adding a unit test.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2247743002

Review-Url: https://codereview.chromium.org/2247743002

Build command buffer with is_component_build=false

BUG=skia:5630
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2265913002

Review-Url: https://codereview.chromium.org/2265913002

Adjust angles in circular arcs GMs to test smaller arcs

BUG=skia:5227
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2267713002

Review-Url: https://codereview.chromium.org/2267713002

add SkMatrixPriv for specialized helpers

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2268443002

Review-Url: https://codereview.chromium.org/2268443002