skia: Add support for CHROMIUM_image backed textures.

I created a new abstract base class TextureStorageAllocator that consumers of
Skia can subclass and pass back to Skia. When a surface is created with a
pointer to a TextureStorageAllocator, any textures it creates, or that are
derived from the original surface, will allocate and deallocate storage using
the methods on TextureStorageAllocator.

BUG=https://code.google.com/p/chromium/issues/detail?id=579664
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1623653002

Committed: https://skia.googlesource.com/skia/+/92098e691f10a010e7421125ba4d44c02506bb55

Review URL: https://codereview.chromium.org/1623653002

add kRGBA_F16_SkColorType

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666343002

Review URL: https://codereview.chromium.org/1666343002

Add GL indirect drawing APIs

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666803002

Review URL: https://codereview.chromium.org/1666803002

Add support for GL_EXT_raster_multisample

Updates the GrGLGpu to call glRasterSamplesEXT when multisample is
enabled for a mixed sampled render target, but the stencil test is not
active.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666563003

Review URL: https://codereview.chromium.org/1666563003

Added JSON support for TextBlob et al.
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1674643002

Review URL: https://codereview.chromium.org/1674643002

Optimize the SkRawStream when the input is an asset stream

BUG=b/26841494
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1645963002

Review URL: https://codereview.chromium.org/1645963002

Remove deferred clear from SkGpuDevice

Add combining to GrClearBatch

Fix issue with state tracking in GrGLGpu::createTestingOnlyBackendTexture

Add tests for clearing GPU SkSurfaces and add tests for GrDrawContext::clear().

Add comment that SkCanvas::flush will resolve the RT in the GPU case.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1658823002

Review URL: https://codereview.chromium.org/1658823002

Last bit of GrGLSLFragmentProcessor-derived class cleanup

I believe these are the last two that used the dispreferred constructor pattern.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1667953005

Review URL: https://codereview.chromium.org/1667953005

Move Google3-specific stack limitation logic to template classes.

Remove #ifdefs in other files.

Reapplies https://codereview.chromium.org/1656143003; removing the implicit constructors for GLPtr and GLPtrAlias resolves the build issue on Android.

Also reverts https://codereview.chromium.org/1663013004

Does not change the public API.

TBR=reed
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666203002

Review URL: https://codereview.chromium.org/1666203002

dont upload to Gold from MSAN bot

NOTRY=true
NOTREECHECKS=true
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1673643002

Review URL: https://codereview.chromium.org/1673643002

update CQ with new bot names

NOTRY=true
NOTREECHECKS=true

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1675643002

Review URL: https://codereview.chromium.org/1675643002

Revert of skia: Add support for CHROMIUM_image backed textures. (patchset #14 id:260001 of https://codereview.chromium.org/1623653002/ )

Reason for revert:
Seems to causing the DEPS roll to fail:

   https://codereview.chromium.org/1668293002/

Original issue's description:
> skia: Add support for CHROMIUM_image backed textures.
>
> I created a new abstract base class TextureStorageAllocator that consumers of
> Skia can subclass and pass back to Skia. When a surface is created with a
> pointer to a TextureStorageAllocator, any textures it creates, or that are
> derived from the original surface, will allocate and deallocate storage using
> the methods on TextureStorageAllocator.
>
> BUG=https://code.google.com/p/chromium/issues/detail?id=579664
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1623653002
>
> Committed: https://skia.googlesource.com/skia/+/92098e691f10a010e7421125ba4d44c02506bb55

TBR=bsalomon@chromium.org,cblume@chromium.org,bsalomon@google.com,robertphillips@google.com,egdaniel@google.com,reed@google.com,erikchen@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=https://code.google.com/p/chromium/issues/detail?id=579664

Review URL: https://codereview.chromium.org/1669213002

Set sRGB flag for PNGs with an sRGB chunk

BUG=skia:3471
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1667823004

Review URL: https://codereview.chromium.org/1667823004

Fix memory leak in SkSpecialSurface

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666373002

TBR=jcgregorio@google.com
NOTRY=true

Review URL: https://codereview.chromium.org/1666373002

Add SkAutoPixmapStorage::detachPixelsAsData()

Allows passing pixels ownership to clients.

BUG=skia:4896
R=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662353002

Review URL: https://codereview.chromium.org/1662353002

skia: Add support for CHROMIUM_image backed textures.

I created a new abstract base class TextureStorageAllocator that consumers of
Skia can subclass and pass back to Skia. When a surface is created with a
pointer to a TextureStorageAllocator, any textures it creates, or that are
derived from the original surface, will allocate and deallocate storage using
the methods on TextureStorageAllocator.

BUG=https://code.google.com/p/chromium/issues/detail?id=579664
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1623653002

Review URL: https://codereview.chromium.org/1623653002

extend compose-colorfilter to 4f

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1664663003

Review URL: https://codereview.chromium.org/1664663003

speculatively comment out chrome mac perf test crasher

The Chrome Mac perf buildbot is mysteriously crashing.
Maybe it's related to this line of code.

Try taking it out to see if the crash goes away.

R=bsalomon@google.com
BUG=580313
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663673006

Review URL: https://codereview.chromium.org/1663673006

Fixed references to MakeMatrix function
TBR=joshualitt@chromium.org
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666213002

Review URL: https://codereview.chromium.org/1666213002

Add SkSpecialImage & SkSpecialSurface classes

Initial classes.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1579323002

Review URL: https://codereview.chromium.org/1579323002

JSON API cleanups
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1664263002

Review URL: https://codereview.chromium.org/1664263002

Don't print dialog box prior to crashing.

This returns us to our previous behavior when we used 0xbbadbeef
to cause the process to terminate.

BUG=skia:4893
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1667113002

Review URL: https://codereview.chromium.org/1667113002

Improved support for images/bitmaps in SkJSONCanvas
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662063003

Review URL: https://codereview.chromium.org/1662063003

Add SkAndroidCodec::getPixels

This is a synonym for the version of getAndroidPixels that accepts
only three parameters (i.e. no AndroidOptions). It is very similar
to SkCodec::getPixels, so I think the motivation for naming the
version with options differently does not apply here.

Add comments to the header describing defaults.

Update the test to use a template, and delete a lot of redundant code.

Rename a variable to stop shadowing another variable.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1647153002

Review URL: https://codereview.chromium.org/1647153002

Next round of GrGLSLFragmentProcessor-derived class cleanup

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1661143003

Review URL: https://codereview.chromium.org/1661143003

Modify skia serve to return clip rect and view matrix

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662003002

Review URL: https://codereview.chromium.org/1662003002

don't get dismayed by negative scales for HQ

needs https://codereview.chromium.org/1663793003/ to land first

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1668033002

Review URL: https://codereview.chromium.org/1668033002

add note about security and new subclasses

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1667663004

Review URL: https://codereview.chromium.org/1667663004

Revert of Work around exasperating Google3 Android build error. (patchset #1 id:1 of https://codereview.chromium.org/1649493005/ )

Reason for revert:
This workaround no longer seems to be necessary.

Original issue's description:
> Work around exasperating Google3 Android build error.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1649493005
>
> Committed: https://skia.googlesource.com/skia/+/900901fb1ba5170549670829fc8f9d9b590ff5f0

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

Review URL: https://codereview.chromium.org/1669663003

Fix Unpremul Index8 tests

TBR=scroggo@google.com

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663323002

Review URL: https://codereview.chromium.org/1663323002

SSE optimizations for GrayAlpha -> RGBA/BGRA Premul/Unpremul

Swizzle Runtime (Dell Venue 8)
Unpremul 0.17x
Premul   0.20x

PNG Decode Runtime on GrayAlpha Encoded PNGs (Dell Venue 8)
Unpremul Regular  0.91x
Unpremul ZeroInit 0.92x
Premul   Regular  0.84x
Premul   ZeroInit 0.85x

BUG=skia:4767
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666853002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1666853002

Make SkImage::readYUV8Planes const

TBR=reed@google.com

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1660393003
BUG=skia:

Review URL: https://codereview.chromium.org/1660393003 .

Increase max vertex attrib count in GrPrimitiveProcessor

Increases the max vertex attrib count in GrPrimitiveProcessor from 6 to
8. This is the required minimum by ES 2.0 for GL_MAX_VERTEX_ATTRIBS.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663513004

Review URL: https://codereview.chromium.org/1663513004

SkMojo: generate on Darwin/x86_64

also clean up generate and skmojo.gyp to work right.

BUG=skia:4891
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1660043004

Review URL: https://codereview.chromium.org/1660043004

Fix assert accessing pipeline from initBatchTracker

Marks fPipelineInstalled as true before calling initBatchTracker. The
batch subclass ought to be able to access the newly constructed
pipeline object during its init.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666813002

Review URL: https://codereview.chromium.org/1666813002

Fix bugs in SkBmpCodec

The decode should not depend on the requested alpha type.

These were exposed by:
https://codereview.chromium.org/1641273003

This should cause the number of untriaged images in Gold
to go to zero.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663303002

Review URL: https://codereview.chromium.org/1663303002

Revert "Factor out xcode c/cpp settings into variables to apply to cmake builds."

This reverts commit 32235eac4b5c625e639757b8247f6e64ce4dc4c8.
This change should no longer be necessary now that gyp has been updated.

Review URL: https://codereview.chromium.org/1664923002

Test unpremul decodes

This is a workaround until SkCanvas supports drawing kUnpremul sources.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1661913005

Review URL: https://codereview.chromium.org/1661913005

Clean up GrGLSLFragmentProcessor-derived classes

In some other patches it was observed that many of the GrGLSLFragmentProcessor-derived classes needlessly pass a parameter to their constructors.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666773002

Review URL: https://codereview.chromium.org/1666773002

Reduce stack usage for Google3.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663013004

Review URL: https://codereview.chromium.org/1663013004

Treat bad values passed to --images as a fatal error

If an option is passed to --images that is either a non-existent path or
a folder with no images matching the supported types, assume this is
an error and exit, so they can supply a valid path instead.

Share code between DM and nanobench in SkCommonFlags.

nanobench now behaves more like DM - it will check a directory for
images that match the supported extensions.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1611323004

Review URL: https://codereview.chromium.org/1611323004

SkMojo: test linking Skia against the Mojo SDK

TODO: build on systems other than Linux.

Add mojo_skd to the DEPS.

Add a DM::Via called `mojo-`.

everything is hidden behind the gyp variable `skia_mojo`.
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1644043003

Review URL: https://codereview.chromium.org/1644043003

Roll third_party/externals/gyp 08429da795..87ac4d0e63fc (63 commits).

https://chromium.googlesource.com/external/gyp.git/+log/08429da795..87ac4d0e63fc

This is primarily to pick up changes to the CMake generator for Mac.

Review URL: https://codereview.chromium.org/1669503003

DM: add uninteresting hashes FYI

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663223002

Review URL: https://codereview.chromium.org/1663223002

NEON optimizations for GrayAlpha -> RGBA/BGRA Premul/Unpremul

PNG Decode Time Nexus 6P (for a test set of GrayAlpha encoded PNGs)
Regular   Unpremul 0.91x
Zero Init Unpremul 0.92x
Regular   Premul   0.84x
Zero Init Premul   0.86x

BUG=skia:4767
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663623002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1663623002

Add an avoidXfermode GM to exercise coverage
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663133003

Review URL: https://codereview.chromium.org/1663133003

Handle gray alpha conversions in SkSwizzler

Nothing fancy here.  By doing the entire conversion ourselves, we
only need to make one conversion pass over each row.  Additionally,
we optimize the premultiply since we know each color component of
the pixel is identical.

This will also enable us to follow up with platform specific
optimizations.

PNG Decode Time Nexus 6P (for a test set of GrayAlpha encoded PNGs)
Regular   Unpremul 0.95x
Zero Init Unpremul 0.94x
Regular   Premul   0.91x
Zero Init Premul   0.90x

BUG=skia:4767
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1665583002

Review URL: https://codereview.chromium.org/1665583002

Add --benchType and --sourceType to nanobench.

To run only codec benches:
   $ out/Release/nanobench --images <> --benchType skcodec

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666743002

Review URL: https://codereview.chromium.org/1666743002

Rename SkBitmapProcStateAutoMapper methods

x(),y() -> fractionalIntX(), fractionalIntY()

(to clarify the return type)

Also add fixed & int helpers.

R=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1666433003
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1666433003

use left-shift macro to fix/avoid undefined behvarior warnings

motivated by https://codereview.chromium.org/1657333002/

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1661873003

Review URL: https://codereview.chromium.org/1661873003

optimize src mode (opaque src in srcover) singleton with aa (e.g. a8 text mask)

Before:
   8/8   MB 1 9.09ms 10.8ms 10.3ms 11.5ms 9% █▆▆▁▁▁▂▅█▆ nonrendering xfer4f_srcover_aa_1_opaque_linear
   8/8   MB 1 10.2ms 12.1ms 11.7ms 13.2ms 9% ▅▇▁▂▁▄█▆▅▆ nonrendering xfer4f_srcover_aa_1_opaque_srgb

After:
   8/8   MB 1 1.6ms 1.68ms 1.73ms 2.17ms 10% ▄▄█▁▃▁▁▂▁▁ nonrendering xfer4f_srcover_aa_1_opaque_linear
   8/8   MB 1 3.13ms 3.62ms 3.97ms 5.81ms 21% █▃▁▂▆▂▂▂▃▂ nonrendering xfer4f_srcover_aa_1_opaque_srgb

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1664713003

Review URL: https://codereview.chromium.org/1664713003

Make helper function on GrGLSLProgramDataManager non virtual

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663833002

Review URL: https://codereview.chromium.org/1663833002

Disable RAW decodes on Windows

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662643003

Review URL: https://codereview.chromium.org/1662643003

extend gm to test aa[] parameter on xfer4f procs

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663643002

Review URL: https://codereview.chromium.org/1663643002

Add gpu implementation of SkAvoidXfermode
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1658623002

TBR=bsalomon@google.com

Committed: https://skia.googlesource.com/skia/+/15691a055db9b68c9b48f589e48d8a85888cf83f

Review URL: https://codereview.chromium.org/1658623002

Support decoding opaque to *premul

If a client requests unpremul or premul from an opaque SkCodec,
support it. The opaque image can be treated as any of them, though
it will be less efficient to draw than if the client had used
opaque.

Change the filling code (i.e. for incomplete images) to base its color on
the source alpha type. Prior to adding the support to decode opaque to
any, it was fine to use either source or dest (which would have yielded
the same result). If the client requests non-opaque, we do not want this
to switch the fill value from black to transparent. This also allows
simplifying the signatures for getFillValue and onGetFillValue.

In CodexTest, expect the same result when decoding opaque to *premul,
and compare to the opaque version.

BUG=skia:4616
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1641273003

Review URL: https://codereview.chromium.org/1641273003

Hack together MSAN build.

  - Build our own: freetype.
  - Avoid using: fontconfig, expat, GPU drivers.
  - Lie about safety: libjpeg (only from src/images... need to in src/codec?)

To run:
  $ tools/xsan_build memory dm
  $ out/Debug/dm -v --match ~Codec ~BlurLargeImage ~FontMgrAndroidParser

Notes:
  - Codec triggers issues in piex.
  - BlurLargeImage probably has bugs in the GM only.
  - FontMgrAndroidParser uses expat.

BUG=skia:4550
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1665823002

NOTREECHECKS=true

Review URL: https://codereview.chromium.org/1665823002

Revert of Test CodecImageGenerator on GPU bots (patchset #1 id:1 of https://codereview.chromium.org/1663453002/ )

Reason for revert:
Failures:
D/skia    ( 3581):  gpu image gen frame_larger_than_image.gif: Could not create a surface.
D/skia    ( 3581):  gpu image gen offsets_too_large.gif: Could not create a surface.
D/skia    ( 3581):  gpu image gen PANO_20121023_214540.jpg: Could not create a surface.
D/skia    ( 3581):  gpu image gen interlaced1.png: Could not create a surface.
D/skia    ( 3581):  gpu image gen interlaced2.png: Could not create a surface.
D/skia    ( 3581):  gpu image gen interlaced3.png: Could not create a surface.
D/skia    ( 3581):  gpu image gen Canon_5D2.dng: Could not create a surface.
D/skia    ( 3581):  gpu image gen Fuji_X20.dng: Could not create a surface.
D/skia    ( 3581):  gpu image gen HTC.dng: Could not create a surface.
D/skia    ( 3581):  gpu image gen lg_g4_iso_800.dng: Could not create a surface.
D/skia    ( 3581):  gpu image gen Canon_G7X.CR2: Could not create a surface.
D/skia    ( 3581):  gpu image gen Pentax_K5.DNG: Could not create a surface.
D/skia    ( 3581):  gpu image gen Nikon_1AW1.NEF: Could not create a surface.
D/skia    ( 3581):  gpu image gen Nikon_1J4.NEF: Could not create a surface.
D/skia    ( 3581):  gpu image gen Nikon_P330.NRW: Could not create a surface.
D/skia    ( 3581):  gpu image gen Olympus_E-PL3.ORF: Could not create a surface.
D/skia    ( 3581):  gpu image gen Olympus_PL7.ORF: Could not create a surface.
D/skia    ( 3581):  gpu image gen Pentax_K5.PEF: Could not create a surface.
D/skia    ( 3581):  gpu image gen Samsung_NX3000.SRW: Could not create a surface.

Original issue's description:
> Test CodecImageGenerator on GPU bots
>
> In crrev.com/1549473003, CodecImageGenerator implemented getYUV8Planes,
> so that we can test on a GPU bot. Update the arguments to DM so that
> we run CodecImageGenerator on GPU bots.
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663453002
>
> Committed: https://skia.googlesource.com/skia/+/c9715406c4c9c995e5661a4ea2188fb8643845c0

TBR=msarett@google.com,scroggo@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/1663143002

Add SkMSAN.h

This lets us tag up pieces of code as requiring initialized inputs.

Almost all code requires initialized inputs, of course.  This is for
code that works correctly with uninitialized data but triggers false
positive warnings in MSAN.  E.g., imagine MSAN's found use of uninitialized
data in this max function:

  static uint8_t max(uint8_t x, uint8_t y) { return x > y ? x : y; }

There's no bug in here... if there's uninitialized data being branched upon
here for the first time, it's sure not max's fault, it's its caller's fault.

So we might do this:
  static uint8_t max(uint8_t x, uint8_t y) {
      // This function uses branching, so if MSAN finds a problem here,
      // we can assert x and y are initialized.  This will remind us the
      // problem somewhere in the caller or above, not here.
      sk_msan_assert_initialized(&x, &x+1);
      sk_masn_assert_initialized(&y, &y+1);
      return x > y ? x : y;
  }

By allowing code to assert its inputs must be initialized,
we can make the blame for use of uninitialized data more clear.

(Sometimes we have another option, to rewrite the code to avoid branching:
  static uint8_t max(uint8_t x, uint8_t y) {
      // This function is branchfree, so MSAN won't complain here.
      // No real need to assert anything as requiring initialization.
      int diff = x - y;
      int negative = diff >> (sizeof(int)*8 - 1);
      return (y & negative) | (x & ~negative);
  }
These approaches to fixing MSAN false positives are orthogonal.)

BUG=chromium:574114
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1658913005
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1658913005

Add option to specify a font collection when creating a
DirectWrite font manager.

The corresponding Chromium change can be found at
https://codereview.chromium.org/1591883002/ .

TBR=reed
This is a trivial and long planned addition to the API.

Review URL: https://codereview.chromium.org/1607083003

Do not upload to trusted partner's bucket if it is a dry run

BUG=skia:3751
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1665803002

# Trybots do not exercise this code.
NOTRY=true

Review URL: https://codereview.chromium.org/1665803002

Always build our own zlib.

If we want to have an MSAN build, it'll help if we can build our own zlib
so that it's instrumented by MSAN.

Today we build our own zlib on Windows, but require the system to provide it
elsewhere.  This just makes everyone build it (except Android framework of course).

This drops the SIMD files.  They're only used to accelerate deflate
(compression), so they're not terribly interesting to us.  Again, this only
really changes compression speed on Windows bots... pretty niche.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1665843002

Review URL: https://codereview.chromium.org/1665843002

Revert of Add gpu implementation of SkAvoidXfermode (patchset #10 id:170001 of https://codereview.chromium.org/1658623002/ )

Reason for revert:
shader compilation failures

error C1008: undefined variable "null"

https://uberchromegw.corp.google.com/i/client.skia/builders/Test-Ubuntu-GCC-ShuttleA-GPU-GTX660-x86_64-Debug/builds/4064/steps/dm/logs/stdio

Original issue's description:
> Add gpu implementation of SkAvoidXfermode
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1658623002
>
> TBR=bsalomon@google.com
>
> Committed: https://skia.googlesource.com/skia/+/15691a055db9b68c9b48f589e48d8a85888cf83f

TBR=egdaniel@google.com,bsalomon@google.com,robertphillips@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/1663093002

Test CodecImageGenerator on GPU bots

In crrev.com/1549473003, CodecImageGenerator implemented getYUV8Planes,
so that we can test on a GPU bot. Update the arguments to DM so that
we run CodecImageGenerator on GPU bots.
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1663453002

Review URL: https://codereview.chromium.org/1663453002

Fix delete_path in webpages_playback.py

BUG=skia:3751
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662673003

# Trybots do not exercise this code.
NOTRY=true

Review URL: https://codereview.chromium.org/1662673003

Use SkBitmapProcStateAutoMapper for filter samplers also

Observation: filter procs are also biased by s.fFilterOne{X,Y} / 2.  They all do
something along these lines:

  s.fInvProc(s.fInvMatrix,
    SkIntToScalar(x) + SK_ScalarHalf,
    SkIntToScalar(y) + SK_ScalarHalf, &srcPt);

  SkFixed fx = SkScalarToFixed(srcPt.fX) - (s.fFilterOneX >> 1);
  SkFixed fy = SkScalarToFixed(srcPt.fY) - (s.fFilterOneX >> 1);

It's trivial to extend SkBitmapProcStateAutoMapper to handle this internally, and
convert everyone off explicit mapping.

R=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1661613002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1661613002

Add gpu implementation of SkAvoidXfermode
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1658623002

TBR=bsalomon@google.com

Review URL: https://codereview.chromium.org/1658623002

Delete skp dir in partner bucket before uploading

BUG=skia:3751
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1660363002

# Trybots do not exercise this code.
NOTRY=true
TBR=borenet

Review URL: https://codereview.chromium.org/1660363002

Update the dng_sdk version. The new version handles the posix_memalign() better
Remove the '-DNO_POSIX_MEMALIGN'
With this change, most new Android (API>=17) should be able to use posix_memalign()

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1655933002

Review URL: https://codereview.chromium.org/1655933002

Templatize GrNonAtomicRef

Templatizes GrNonAtomicRef so it does not necessarily require a virtual
destructor. This also gives us the flexibility to specialize how
different types get deleted.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1664613002

Review URL: https://codereview.chromium.org/1664613002

Fix for rounded-rect clips with filters.

Don't use the base canvas size to limit raster of complex clips, since
the top canvas size may actually be larger (e.g., a blur filter which
expands the clip bounds to accommodate filter margins). Use the top
canvas bounds instead.

BUG=skia:4879,471212
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1657333002

Review URL: https://codereview.chromium.org/1657333002

Revert of Move Google3-specific stack limitation logic to template classes. Remove #ifdefs in other files. (patchset #2 id:50001 of https://codereview.chromium.org/1656143003/ )

Reason for revert:
See https://codereview.chromium.org/1665603002

Original issue's description:
> Move Google3-specific stack limitation logic to template classes. Remove #ifdefs in other files.
>
> Does not change the public API.
>
> TBR=reed
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1656143003
>
> Committed: https://skia.googlesource.com/skia/+/c92159c8250c62cc47b7b63686538d61d54d2835

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

Review URL: https://codereview.chromium.org/1666503002

Recreate SKPs in a single directory for trusted partners.

Context: https://bugs.chromium.org/p/skia/issues/detail?id=3751#c15

BUG=skia:3751
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662693002

# Trybots do not run this code
NOTRY=true
TBR=borenet

Review URL: https://codereview.chromium.org/1662693002

SSSE3 optimizations for gray -> RGBA (or BGRA)

Swizzle Bench Runtime
Dell Venue 8 0.16x
HP z620      0.47x

PNG Decode Time (for test set of gray encoded PNGs)
Dell Venue 8 0.80x
HP z620      0.96x

BUG=skia:4767
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1657393002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1657393002

Modify skiaserve to support downloads

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1657403002

Review URL: https://codereview.chromium.org/1657403002

NEON optimizations for gray -> RGBA (or BGRA) conversions

Swizzle Bench Runtime
Nexus 6P 0.32x
Nexus 9  0.89x

PNG Decode Time (for test set of gray encoded PNGs)
Nexus 6P 0.88x
Nexus 9  0.91x

BUG=skia:4767
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1656383002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1656383002

Move Google3-specific stack limitation logic to template classes. Remove #ifdefs in other files.

Does not change the public API.

TBR=reed
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1656143003

Review URL: https://codereview.chromium.org/1656143003

Update SK_IMAGE_VERSION to test RAW

Bump SK_IMAGE_VERSION to test the images in v2 in GoogleStorage, which
includes the images from v1 plus test images for SkRawCodec.

Only define skia_decodes_raw on platforms that support it, rather than
defining it always and checking additional conditions to determine
whether to support raw. Further, define it and SK_CODEC_DECODES_RAW
for all targets, so we can use the compile flag in other targets.

In DM, exclude the raw extensions if SK_CODEC_DECODES_RAW is not defined.

Blacklist raw extensions on NexusPlayer, which was running out of memory
when running them.

BUG=skia:4829

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1612113002

Review URL: https://codereview.chromium.org/1612113002

Modify skiaserve to support drawTo

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1651403003

Review URL: https://codereview.chromium.org/1651403003

unroll srcover_1 for blending a single color

Before:
curr/maxrss loops min median mean max stddev samples    config bench
   8/8   MB 1 1.59ms 1.82ms 1.89ms 2.59ms 14% ▁█▃▃▃▃▃▃▃▃ nonrendering xfer4f_srcover_1_alpha_linear
   8/8   MB 1 3.25ms 4.25ms 4.16ms 5.87ms 21% ▁▅▂▁▁▄█▄▅▂ nonrendering xfer4f_srcover_1_alpha_srgb

After:
curr/maxrss loops min median mean max stddev samples    config bench
   8/8   MB 1 915µs 915µs 946µs 1.02ms 4% █▄▇▁▁▁▆▁▁▁ nonrendering xfer4f_srcover_1_alpha_linear
   8/8   MB 1 2.69ms 3.08ms 3.03ms 3.63ms 10% ▁▃▂▁▁█▄▄▄▆ nonrendering xfer4f_srcover_1_alpha_srgb

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1653943002

Review URL: https://codereview.chromium.org/1653943002

Fix SampleApp instructions in the Windows quick-start

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1661553002
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=1661553002

Review URL: https://codereview.chromium.org/1661553002

Remove SK_SUPPORT_LEGACY_SAMPLER_BIAS

The guard has been removed from Chromium, this is now dead code.

TBR=reed@google.com,caryclark@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1658563002

Review URL: https://codereview.chromium.org/1658563002

Sampler bias for all nofilter flavors

Instead of a piecemeal approach to sampler bias (http://crrev.com/1642273002,
http://crrev.com/1539083002), audit all nofilter procs (grep "fInvProc(")
and convert them to SkBitmapProcStateAutoMapper.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1656833002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
R=reed@google.com

Review URL: https://codereview.chromium.org/1656833002

increase inner-loop for xfer4 benches to reduce variance

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1653143004

TBR=

Review URL: https://codereview.chromium.org/1653143004

support for more features when rendering to/from JSON
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662503003

Review URL: https://codereview.chromium.org/1662503003

Fix printfs for Windows and undo speculative test disable

TBR=robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1662533002

Review URL: https://codereview.chromium.org/1662533002 .

Avoid too small reads to bufferMoreData()

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1644893003

Committed: https://skia.googlesource.com/skia/+/fd918e8c739065fa467cf6614181d3a2c5dcadcb

Review URL: https://codereview.chromium.org/1644893003

fix skiaserve crash

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1659703004

Review URL: https://codereview.chromium.org/1659703004

cleanup of skia serve url handling

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1653203002

Review URL: https://codereview.chromium.org/1653203002

Fix fuzzer-found deserialization bug in SkPathRef

This fixes a bug in SkPathRef::CreateFromBuffer found by
fuzzing SkPaintImageFilter.

BUG=582705
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1653003004

Review URL: https://codereview.chromium.org/1653003004

Revert of Avoid too small reads to bufferMoreData() (patchset #4 id:60001 of https://codereview.chromium.org/1644893003/ )

Reason for revert:
Crashed both Nexus 5s:

  https://build.chromium.org/p/client.skia.android/builders/Perf-Android-GCC-Nexus5-CPU-NEON-Arm7-Release/builds/3355
https://build.chromium.org/p/client.skia.android/builders/Perf-Android-GCC-Nexus5-CPU-NEON-Arm7-Release/builds/3354

Original issue's description:
> Avoid too small reads to bufferMoreData()
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1644893003
>
> Committed: https://skia.googlesource.com/skia/+/fd918e8c739065fa467cf6614181d3a2c5dcadcb

TBR=msarett@google.com,scroggo@google.com,yujieqin@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/1656303002

Avoid too small reads to bufferMoreData()

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1644893003

Review URL: https://codereview.chromium.org/1644893003

Speculative fix for windows builders

TBR=robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1651303002

Review URL: https://codereview.chromium.org/1651303002 .

Disable test that is breaking Android and Windows bots.
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1654283002
BUG=skia:

R=robertphillips@google.com

Review URL: https://codereview.chromium.org/1654283002 .

Add ability to extract YUV planes from SkImage

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1513393002

Review URL: https://codereview.chromium.org/1513393002

Fix GL readback code to handle rowbytes correctly for non-32bit formats

Update tests to exercise more rowbytes.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1645043006

Review URL: https://codereview.chromium.org/1645043006

Look beyond SSE2 for Paeth

You can break this CL down into three steps.  Steps 2 and 3 depend on 1.

    Step 1: go to a 16-bit impl.  Speed ~unaffected.
    Step 2: use SSSE3 16-bit abs.  ~20% speedup to Paeth.
    Step 3: use SSE4.1 blendv, total ~25% speedup to Paeth.

Overall this can improve PNG decoding by around 8% end-to-end.

I would feel most comfortable landing this only after we have a bot exercising the SSE4.1 code, either by moving this stuff behind a function pointer (simulating Chrome/Clank) or by adding a builder with at least SSE4.1 at compile time (simulating an Android system build).  We've got plenty of bots building with SSSE3 at compile time to test that path.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dimage&master=false&issue=1657503002

Review URL: https://codereview.chromium.org/1657503002

Move SkColorMatrixFilter implementation to core.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1648933002

Review URL: https://codereview.chromium.org/1648933002

added a default constructor for GrBatchToXPOverrides
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1657803002

Review URL: https://codereview.chromium.org/1657803002

Fix zero-sized blur with crop rect.

Neither the GPU nor CPU paths were correctly handling the crop rect
in this case.

NOTE: this change adds a new test case to the imageblurcropped GM, so
it will have to be rebaselined.

BUG=skia:4876
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1657773002

Review URL: https://codereview.chromium.org/1657773002