Remove transitional explicit operator bool.

These should no longer be necessary. Android's STL should now
be sane.

TBR=reed
This just removes code.

Review URL: https://codereview.chromium.org/1817153002

Make the debug interface a GrGLTestInterface subclass.

The debug interface subsumes the GrDebugGL singleton which allows multiple instances.

Fixes a few issues that existed before this CL (pack alignment wasn't tracked or respected, and we weren't initializing a texture id to 0 in one GM). Apparently this is not currently run on the bots.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1812323002

Review URL: https://codereview.chromium.org/1812323002

Swarming: Ensure depot_tools is in PATH when compiling

Turns out we were "lucky" for our Linux bots which already have depot_tools in PATH. The Mac bot does have depot_tools, but it's not in PATH.  Add it in the script to ensure we can always get to it.

NOTRY=true
BUG=skia:4763
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1823653002

Review URL: https://codereview.chromium.org/1823653002

Make SkComposeImageFilter::onFilterImage filter the bounds given to the inner filter.

Previously, the bounds requested by the caller would be passed to both the
inner and outer filter, but since the outer filter may move pixels, this is not
necessarily sufficient to supply the pixels required by the outer filter to fill
the bounds.

Unit test included.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813373002

Review URL: https://codereview.chromium.org/1813373002

add SK_SUPPORT_LEGACY_PATHEFFECT_PTR for google3 roll fix

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1817113002

TBR=mtklein
NOTRY=True

Review URL: https://codereview.chromium.org/1817113002

Correct comment to refer to SkImageFilter::kReverse_MapDirection.

kBackward_MapDirection does not exist; might as well have the comment exactly
match the valid enumerator.

TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1820883002

Review URL: https://codereview.chromium.org/1820883002

Parse icc profiles and exif orientation from jpeg markers

New resources should be fine to add since they are already
checked into chromium.

BUG=skia:3834
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813273002

Review URL: https://codereview.chromium.org/1813273002

Add note about Chromium's use of old libstdc++ to docs.

Chromium must still build against libstdc++4.6.4, which presents limits
on at least type traits and perhaps performance.

NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=1820873002

Review URL: https://codereview.chromium.org/1820873002

nanobench: Reduce the total number of statistics being captured.

BUG=skia:5092
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1819863002

Review URL: https://codereview.chromium.org/1819863002

Make FreeType usable on Windows.

Add missing functionality so that pdfium can use FreeType
from Windows.

Add an empty custom font manager so pdfium can use FreeType
without any local fonts.

R=bungeman@google.com

Review URL: https://codereview.chromium.org/1817633002

sRGB support in Ganesh. Several pieces:

sRGB support now also requires GL_EXT_texture_sRGB_decode, which allows
us to disable sRGB -> Linear conversion when reading textures. This gives
us an easy way to support "legacy" L32 mode. We disable decoding based on
the pixel config of the render target. Textures can override that behavior
(specifically for format-conversion draws where we want that behavior).

Added sBGRA pixel config, which is not-really-a-format. It's just sRGBA
internally, and the external format is BGR order, so TexImage calls will
swizzle correctly. This lets us interact with sRGB raster surfaces on BGR
platforms.

Devices without sRGB support behave like they always have: conversion from
color type and profile type ignores sRGB and always returns linear pixel
configs.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1789663002

Review URL: https://codereview.chromium.org/1789663002

Fix for unaligned automatic variable address in SSE routines. This used to work on linux but crashed on windows when rendering antialiased glyphs. I tracked the problem down to an alignment issue.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1817813002

Review URL: https://codereview.chromium.org/1817813002

SkPDF: metadata first

Motivation: while experimenting with serializing images immediately, I would like to keep serializing metadata first.

TBR=djsollen@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1780463008

Review URL: https://codereview.chromium.org/1780463008

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=
NO_MERGE_BUILDS
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1817833002

Review URL: https://codereview.chromium.org/1817833002

Don't convert DWORD to int.

DWORD is unsigned, and might not be an int. vs2015u2 complains.

BUG=skia:4553

TBR=mtklein
This is pretty trivial and makes the vs2015 bot compile.

Review URL: https://codereview.chromium.org/1818713002

Templatize SkToXXX.

Makes the checked cast in debug more correct, avoiding new
warnings in vs2015.

BUG=skia:4553

Committed: https://skia.googlesource.com/skia/+/0be9e806af72b3e029e691eef5c891c90d3fd320

Review URL: https://codereview.chromium.org/1814153003

Fix uninitialized fCurrProgramID in GrGLCreateNullInterface

TBR=bsalomon@google.com,robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1817793002

Review URL: https://codereview.chromium.org/1817793002

Document BitmapStateAutoMapper.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1815373002

Review URL: https://codereview.chromium.org/1815373002

disable SK_SUPPORT_LEGACY_PATHEFFECT_PTR for skia builds

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1820593002

TBR=

Review URL: https://codereview.chromium.org/1820593002

Remove uses of SkImageDecoder from samplecode

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1812323003

Review URL: https://codereview.chromium.org/1812323003

Replace uses of SkImageDecoder in src/animator

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1818563002

Review URL: https://codereview.chromium.org/1818563002

Add DescriptorPool and set manager to GrVkProgram

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1765923002

Review URL: https://codereview.chromium.org/1765923002

Fix ANGLE build
TBR=
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1814233003

Review URL: https://codereview.chromium.org/1814233003

Use dynamic state from vulkan pipelines

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813913002

Review URL: https://codereview.chromium.org/1813913002

allow more options for shader blitprocs

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1810383004

Review URL: https://codereview.chromium.org/1810383004

Enable extension support and debug layer.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1785813002
TBR=bsalomon@google.com

Review URL: https://codereview.chromium.org/1785813002

Remove uses of SkImageDecoder from gms

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1791583002

Committed: https://skia.googlesource.com/skia/+/d427266a58af7e6ea6c12c9cd56ade3e179a0c04

Review URL: https://codereview.chromium.org/1791583002

Store null GL context's state in interface object

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1814113002

Review URL: https://codereview.chromium.org/1814113002

Implement Vulkan GrBackendObject for textures.

BUG=skia:5043
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808263002

Review URL: https://codereview.chromium.org/1808263002

Revert of Templatize SkToXXX. (patchset #2 id:20001 of https://codereview.chromium.org/1814153003/ )

Reason for revert:
Chrome does not yet have std::underlying_type.

Original issue's description:
> Templatize SkToXXX.
>
> Makes the checked cast in debug more correct, avoiding new
> warnings in vs2015.
>
> BUG=skia:4553
>
> Committed: https://skia.googlesource.com/skia/+/0be9e806af72b3e029e691eef5c891c90d3fd320

TBR=reed@google.com,mtklein@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4553

Review URL: https://codereview.chromium.org/1820433002

Propogate SkBudgeted for NewFromDeferredTextureImageData

The budgeted flag was not propogated from
NewFromDeferredTextureImageData to MakeTextureFromPixmap, resulting in
the created textures always being budgeted, even when SkBudgeted::kNo
was passed in.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1809663002

Review URL: https://codereview.chromium.org/1809663002

Remove --verbose from DM on Swarming bots

TBR=mtklein
BUG=skia:4763
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1817563002

Review URL: https://codereview.chromium.org/1817563002

Reland of "switch patheffects over to sk_sp (patchset #5 id:80001 of https://codereview.chromium.org/1813553005/ )"

This reverts commit f28ad894272018fd2855e3f77ea1236ea0cce1c0.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813123003

TBR=

Review URL: https://codereview.chromium.org/1813123003

Templatize SkToXXX.

Makes the checked cast in debug more correct, avoiding new
warnings in vs2015.

BUG=skia:4553

Review URL: https://codereview.chromium.org/1814153003

Fix pessimizing move in SamplePictFile

../../samplecode/SamplePictFile.cpp:194:20: error: moving a local object in a return statement prevents copy elision [-Werror,-Wpessimizing-move]
            return std::move(pic);
                               ^

R=reed@google.com,mtklein@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1809943005

Review URL: https://codereview.chromium.org/1809943005

custom ssse3 srcover_n_srgb_bw, about 1.8x speedup

This is a little demo of the sorts of speedups we can get from working in planar format, or even just a mini-planar of 4 pixels at a time like I'm doing here.

I chose this blit by running
  $ out/Release/nanobench --config srgb --match skp
and looking for the hottest sRGB-related method.
After this CL, src_1 and src_n become hotter than srcover_n.  They can probably get a similar treatment.

We transpose three times in this function:
   - dst after reading, as part of the zero-extension and conversion to float
   - src after reading, _MM_TRANSPOSE4_PS (which expands to 8 cheap instructions)
   - result before writing, the last _mm_shuffle_epi8
If we changed our buffer format to a mini-planar format like rrrr gggg bbbb aaaa, we could eliminate the src transpose and get another small speedup, to right around 2x.

This code leans pretty heavily on SSSE3, so if we want it to speed up Windows+Linux Chrome, it'll eventually want to go behind a function pointer.

This also appears to fix what looks like overflow in a few GMs, most noticeably in hairmodes.  This is something we'd better look into...

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813263002

Review URL: https://codereview.chromium.org/1813263002

4f linear gradient shader blitters

Add F16 specializations to support writing to half-float dests.

Add color profile template arg across the board to support writing
to sRGB dests.

R=reed@google.com,mtklein@google.com,herb@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808963005

Review URL: https://codereview.chromium.org/1808963005

Revert of switch patheffects over to sk_sp (patchset #5 id:80001 of https://codereview.chromium.org/1813553005/ )

Reason for revert:
some build breaks, possibly related to paint having to know what a patheffect is

Original issue's description:
> switch patheffects over to sk_sp
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813553005
>
> Committed: https://skia.googlesource.com/skia/+/9fbee18f691a0afed1e38a851048ce06063505ed

TBR=caryclark@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/1817543002

switch patheffects over to sk_sp

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813553005

Review URL: https://codereview.chromium.org/1813553005

Make Win runtime DLLs writeable before overwriting

This should fix the VS2015 bot which is failing because we can't
overwrite read-only DLL files.

BUG=skia:4553
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813233002

Review URL: https://codereview.chromium.org/1813233002

Add SkSpecialImage::makeTextureImage entry point

This is calved off of: https://codereview.chromium.org/1785643003 (Switch SkBlurImageFilter over to new onFilterImage interface)

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813813002

Committed: https://skia.googlesource.com/skia/+/05849018c85403a34b88819db1c4bcf713b70a2b

Review URL: https://codereview.chromium.org/1813813002

Clean up SSSE3 and SSE4 stubs.

We added these stubs to work around OpenBSD's old compiler, which had
support for SSE2 but not SSSE3 or SSE4.

We now already have other unstubbed files that require SSSE3 and SSE4 compiler
support.  All the compilers we support have SSSE3 and SSE4 support, and all the
way up to at least AVX2.

(Requiring C++11 has had some nice ripple effects...)

And, <immintrin.h> is already auto-included for these files, so no need for smmintrin or tmmintrin.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1810183003
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1810183003

return pictures as sk_sp

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1811703002

Review URL: https://codereview.chromium.org/1811703002

allow one zero length dash

If the constructed stroke that represents a dash has a
single dash of length zero, and the end cap is square or
round, draw the cap.

The old code initialized the initial dash length to zero,
making it ambiguous whether the first length is zero or
not.

R=robertphillips@google.com
BUG=583299
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1805963002

Committed: https://skia.googlesource.com/skia/+/5e1a24808415df2748822e8082e21a361362cdfe

Review URL: https://codereview.chromium.org/1805963002

Revert of Add SkSpecialImage::makeTextureImage entry point & update filterInput (patchset #3 id:40001 of https://codereview.chromium.org/1813813002/ )

Reason for revert:
Suspected cause of layout test css3/filters/effect-reference-tile-hw.htmlto crash.

https://build.chromium.org/p/tryserver.blink/builders/linux_blink_rel/builds/83847/steps/webkit_tests%20%28with%20patch%29/logs/stdio

It could also be the previous change (but it appears this one depends on that one).

Original issue's description:
> Add SkSpecialImage::makeTextureImage entry point
>
> This is calved off of: https://codereview.chromium.org/1785643003 (Switch SkBlurImageFilter over to new onFilterImage interface)
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813813002
>
> Committed: https://skia.googlesource.com/skia/+/05849018c85403a34b88819db1c4bcf713b70a2b

TBR=bsalomon@google.com,senorblanco@google.com,senorblanco@chromium.org,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/1811973005

Revert of allow one zero length dash (patchset #8 id:140001 of https://codereview.chromium.org/1805963002/ )

Reason for revert:
Causes the dash bench to crash.

Example crash:
https://build.chromium.org/p/client.skia/builders/Perf-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release/builds/5581/steps/nanobench/logs/stdio

Original issue's description:
> allow one zero length dash
>
> If the constructed stroke that represents a dash has a
> single dash of length zero, and the end cap is square or
> round, draw the cap.
>
> The old code initialized the initial dash length to zero,
> making it ambiguous whether the first length is zero or
> not.
>
> R=robertphillips@google.com
> BUG=583299
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1805963002
>
> Committed: https://skia.googlesource.com/skia/+/5e1a24808415df2748822e8082e21a361362cdfe

TBR=robertphillips@google.com,reed@google.com,caryclark@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=583299

Review URL: https://codereview.chromium.org/1808303004

allow one zero length dash

If the constructed stroke that represents a dash has a
single dash of length zero, and the end cap is square or
round, draw the cap.

The old code initialized the initial dash length to zero,
making it ambiguous whether the first length is zero or
not.

R=robertphillips@google.com
BUG=583299
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1805963002

Review URL: https://codereview.chromium.org/1805963002

Add SkSpecialImage::makeTextureImage entry point

This is calved off of: https://codereview.chromium.org/1785643003 (Switch SkBlurImageFilter over to new onFilterImage interface)

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813813002

Review URL: https://codereview.chromium.org/1813813002

Switch SkSpecialImage & SkSpecialSurface classes over to smart pointers

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1812023002

Review URL: https://codereview.chromium.org/1812023002

Add SkImageGeneratorWIC

This will be a replacement for SkImageDecoder_WIC.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1785613010
CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Ubuntu-GCC-x86_64-Release-CMake-Trybot,Build-Mac-Clang-x86_64-Release-CMake-Trybot

Review URL: https://codereview.chromium.org/1785613010

documentation: simplify pdf diagram

NOTRY=true
TBR=

Review URL: https://codereview.chromium.org/1810943003

documentation: simplify pdf diagram

NOTRY=true
TBR=

Review URL: https://codereview.chromium.org/1810943003

Revert of sRGB support in Ganesh. Several pieces: (patchset #12 id:220001 of https://codereview.chromium.org/1789663002/ )

Reason for revert:
We're getting sRGB non-8888 configs?

Original issue's description:
> sRGB support in Ganesh. Several pieces:
>
> sRGB support now also requires GL_EXT_texture_sRGB_decode, which allows
> us to disable sRGB -> Linear conversion when reading textures. This gives
> us an easy way to support "legacy" L32 mode. We disable decoding based on
> the pixel config of the render target. Textures can override that behavior
> (specifically for format-conversion draws where we want that behavior).
>
> Added sBGRA pixel config, which is not-really-a-format. It's just sRGBA
> internally, and the external format is BGR order, so TexImage calls will
> swizzle correctly. This lets us interact with sRGB raster surfaces on BGR
> platforms.
>
> Devices without sRGB support behave like they always have: conversion from
> color type and profile type ignores sRGB and always returns linear pixel
> configs.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1789663002
>
> Committed: https://skia.googlesource.com/skia/+/9e3f1bf4e5cd8fc59554f986f36d6b034e99f9eb

TBR=reed@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
BUG=skia:

Review URL: https://codereview.chromium.org/1814533003

Add asRRect method to SkClipStack::Element

Adds an asRRect method alongside asPath, for clip implementations that
can be generalized to round rects.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1793373002

Review URL: https://codereview.chromium.org/1793373002

DM: make --verbose send the vlog to stderr.

See if this works out for you?
I've just turned off --verbose on the bots, so this ought to be safe now.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808203004

Review URL: https://codereview.chromium.org/1808203004

sRGB support in Ganesh. Several pieces:

sRGB support now also requires GL_EXT_texture_sRGB_decode, which allows
us to disable sRGB -> Linear conversion when reading textures. This gives
us an easy way to support "legacy" L32 mode. We disable decoding based on
the pixel config of the render target. Textures can override that behavior
(specifically for format-conversion draws where we want that behavior).

Added sBGRA pixel config, which is not-really-a-format. It's just sRGBA
internally, and the external format is BGR order, so TexImage calls will
swizzle correctly. This lets us interact with sRGB raster surfaces on BGR
platforms.

Devices without sRGB support behave like they always have: conversion from
color type and profile type ignores sRGB and always returns linear pixel
configs.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1789663002

Review URL: https://codereview.chromium.org/1789663002

Fix pointer aliasing bug in SkImageFilter::computeFastBounds.

Since src and dst are explicitly allowed to alias (according to a comment in
SkPaint.h), it is problematic to have the first input filter mutate dst, since
we still need access to the previous value to provide to the other input
filters.

To resolve this, SkImageFilter::computeFastBounds makes a copy of src on the
stack, and passes that to its inputs instead.

Also add a unit test that would have detected the issue.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1800263002

Review URL: https://codereview.chromium.org/1800263002

just write the verbose log file any time we have a --writePath

oughta fix the Android bots

TBR=bungeman@google.com

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813933002

Review URL: https://codereview.chromium.org/1813933002

Update how we send draws to gpu backend to reduce state setting.

The main change here is that we pull primitive type off of the vertices, we set the gpu state on gpu once per pipeline/prim proc draw batch, and we create the ProgramDescriptor only for the Cache/ProgramBuilder.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1806983002

Review URL: https://codereview.chromium.org/1806983002

free -> reset

The C++ standard library uses ".reset()" where we sometimes write ".free()".
We also use ".reset()" quite a lot.  This standardizes on ".reset()".

This is one more step towards dropping SkAutoTDelete in favor of the standard
std::unique_ptr.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1811723002

Committed: https://skia.googlesource.com/skia/+/0e3738db89e86035ed5d4f629bf58b817b1e5274

Review URL: https://codereview.chromium.org/1811723002

update callsites for Make image factories

not forced yet, as we still have the build-guard. waiting on chrome CL

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1810813003

TBR=

Review URL: https://codereview.chromium.org/1810813003

Fix gdb debugging for the latest NDK release.
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1806343003

Review URL: https://codereview.chromium.org/1806343003

Disable bad gif decode on CG

This will save me the trouble of triaging this bad image over and
over again :).

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1811943002

Review URL: https://codereview.chromium.org/1811943002

Revert of Remove uses of SkImageDecoder from gms (patchset #2 id:20001 of https://codereview.chromium.org/1791583002/ )

Reason for revert:
Suspected of causing assertion failures in unrelated tests on Windows.

Original issue's description:
> Remove uses of SkImageDecoder from gms
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1791583002
>
> Committed: https://skia.googlesource.com/skia/+/d427266a58af7e6ea6c12c9cd56ade3e179a0c04

TBR=scroggo@google.com,reed@google.com,msarett@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/1806383002

Use sktarray.begin() instead of &sktarray[0].

Evaluating &sktarray[0] when sktarray.count() == 0 asserts.
This was causing ASAN failures.

TBR=bsalomon
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1812963002

try to flush stdio in the crash handler

NOTREECHECKS=true

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808853004

Review URL: https://codereview.chromium.org/1808853004

Revert of free -> reset (patchset #3 id:40001 of https://codereview.chromium.org/1811723002/ )

Reason for revert:
Suspect for Win10 failures.

Original issue's description:
> free -> reset
>
> The C++ standard library uses ".reset()" where we sometimes write ".free()".
> We also use ".reset()" quite a lot.  This standardizes on ".reset()".
>
> This is one more step towards dropping SkAutoTDelete in favor of the standard
> std::unique_ptr.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1811723002
>
> Committed: https://skia.googlesource.com/skia/+/0e3738db89e86035ed5d4f629bf58b817b1e5274

TBR=reed@google.com,mtklein@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/1813843002

Fix Linux Swarming bots

NOTRY=true
NOTREECHECKS=true
BUG=skia:4763
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1806323003

Review URL: https://codereview.chromium.org/1806323003

Some fixes for Swarming bots

BUG=skia:4763
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1813443003

Review URL: https://codereview.chromium.org/1813443003

Allow SkGpuDevice::drawSprite to handle subset SkBitmaps

With the upcoming switch to SkSpecialImage in the blur image filter, it is possible for drawSprite calls to use an SkBitmap with an offset (e.g., a blur image filter with a cropRect and 0 blur radii). This simply updates the drawSprite code path to be origin aware.

This is calved off of: https://codereview.chromium.org/1785643003/ (Switch SkBlurImageFilter over to new onFilterImage interface)

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808743003

Review URL: https://codereview.chromium.org/1808743003

Allow const& for SkImages and SkPictures in draw methods.

This is just a convenience for callers, since we are (strongly) encouraging them to use sk_sp<> for managing image objects (and pictures btw). No change under the hood, as we are keeping our agnostic approach to these for the protected virtuals (since often we don't want to ref the parameter).

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1777403002

Review URL: https://codereview.chromium.org/1777403002

Remove uses of SkImageDecoder from gms

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1791583002

Review URL: https://codereview.chromium.org/1791583002

Remove uses of SkImageDecoder from skdiff

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1788643003

Review URL: https://codereview.chromium.org/1788643003

make sausages rounder and portable

TBR=bungeman@google.com
BUG=226341
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808203003

Review URL: https://codereview.chromium.org/1808203003

Add SkSpecialImage::extractSubset & NewFromPixmap

This is calved off of: https://codereview.chromium.org/1785643003/ (Switch SkBlurImageFilter over to new onFilterImage interface)

This now relies on: https://codereview.chromium.org/1813483002/ (ImagePixelLocker now manually allocates SkPixmap) to clean up the uses of SkAutoPixmapStorage in Chromium

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1787883002

Committed: https://skia.googlesource.com/skia/+/250581493a0859987e482810879e85e5ac2dc002

Review URL: https://codereview.chromium.org/1787883002

Remove SkipZeroesBench

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1784333003

Review URL: https://codereview.chromium.org/1784333003

free -> reset

The C++ standard library uses ".reset()" where we sometimes write ".free()".
We also use ".reset()" quite a lot.  This standardizes on ".reset()".

This is one more step towards dropping SkAutoTDelete in favor of the standard
std::unique_ptr.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1811723002

Review URL: https://codereview.chromium.org/1811723002

don't create zero length intervals

Dashing a pattern without zero-length intervals should
not create them if the end of the on interval coincides
with the beginning of the initial dash offset.

R=reed@google.com
BUG=591993
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1766243004

Committed: https://skia.googlesource.com/skia/+/18bbd00190623fb6cdb119df4a118ac3c1aed52a

Review URL: https://codereview.chromium.org/1766243004

Add back SkAutoTUnref::detach() for Android temporarily.

TBR=reed@google.com

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1812843002

Review URL: https://codereview.chromium.org/1812843002

Add back SkAutoTDelete::detach() for Android temporarily.

TBR=

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1811483004

Review URL: https://codereview.chromium.org/1811483004

Image filters: fix the zero-sigma fast path in SkBlurImageFilter.

The fast path was not correctly taking the input srcOffset into account.

BUG=595324
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1807673005

Review URL: https://codereview.chromium.org/1807673005

Un-deprecate new methods to fix Android.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808663004

Review URL: https://codereview.chromium.org/1808663004

detach -> release

The C++ standard library uses the name "release" for the operation we call "detach".

Rewriting each "detach(" to "release(" brings us a step closer to using standard library types directly (e.g. std::unique_ptr instead of SkAutoTDelete).

This was a fairly blind transformation.  There may have been unintentional conversions in here, but it's probably for the best to have everything uniformly say "release".

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1809733002

Review URL: https://codereview.chromium.org/1809733002

add blitprocs to shaders

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1793883002

Review URL: https://codereview.chromium.org/1793883002

Minor mingw cross compilation fix.

Windows.h is actually windows.h.

Review URL: https://codereview.chromium.org/1804163005

Revert of Add SkSpecialImage::extractSubset & NewFromPixmap (patchset #5 id:80001 of https://codereview.chromium.org/1787883002/ )

Reason for revert:
Need to wean ImagePixelLocker.h off of SkAutoPixmapStorage :(

Original issue's description:
> Add SkSpecialImage::extractSubset & NewFromPixmap
>
> This is calved off of: https://codereview.chromium.org/1785643003/ (Switch SkBlurImageFilter over to new onFilterImage interface)
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1787883002
>
> Committed: https://skia.googlesource.com/skia/+/250581493a0859987e482810879e85e5ac2dc002

TBR=bsalomon@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/1808833002

Use std::unique_ptr.

TBR=reed@google.com

Committed: https://skia.googlesource.com/skia/+/20c1e3abfc681771f73eb19fde7284196e028940

Committed: https://skia.googlesource.com/skia/+/3dd9ed37c24611af86f0fe374bd3698b63f09450
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1780933003

Committed: https://skia.googlesource.com/skia/+/2691d76a06e1af6282f8b3a3140cc93361be10c4

Review URL: https://codereview.chromium.org/1780933003

GrTessellator: don't reuse the previous vertex buffer on a cache miss.

Just let the cache give us the correct entry for the given size.
This simplifies the code (no mysterious non-const ref member).
I don't see any performance effect from this change, but will watch
the bots closely.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1800623002

Review URL: https://codereview.chromium.org/1800623002

Doc: pdf illustration

TBR=
NOTRY=true

Review URL: https://codereview.chromium.org/1807963002

Revert of Use std::unique_ptr. (patchset #10 id:170001 of https://codereview.chromium.org/1780933003/ )

Reason for revert:
aww jeez, I dropped ps 8...

Original issue's description:
> Use std::unique_ptr.
>
> TBR=reed@google.com
>
> Committed: https://skia.googlesource.com/skia/+/20c1e3abfc681771f73eb19fde7284196e028940
>
> Committed: https://skia.googlesource.com/skia/+/3dd9ed37c24611af86f0fe374bd3698b63f09450
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1780933003
>
> Committed: https://skia.googlesource.com/skia/+/2691d76a06e1af6282f8b3a3140cc93361be10c4

TBR=bungeman@google.com,mtklein@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1809693002

Add SkSpecialImage::extractSubset & NewFromPixmap

This is calved off of: https://codereview.chromium.org/1785643003/ (Switch SkBlurImageFilter over to new onFilterImage interface)

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1787883002

Review URL: https://codereview.chromium.org/1787883002

Cleanup download scripts related to new Android NDK.
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1810603002

Review URL: https://codereview.chromium.org/1810603002

Use std::unique_ptr.

TBR=reed@google.com

Committed: https://skia.googlesource.com/skia/+/20c1e3abfc681771f73eb19fde7284196e028940

Committed: https://skia.googlesource.com/skia/+/3dd9ed37c24611af86f0fe374bd3698b63f09450
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1780933003

Review URL: https://codereview.chromium.org/1780933003

Split uninteresting hash lines correctly on Windows.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1806903002

Review URL: https://codereview.chromium.org/1806903002

Begin tracking GL_TEXTURE_ and GL_DRAW_INDIRECT_ buffer bindings

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1796713005

Review URL: https://codereview.chromium.org/1796713005

pass cap to dash text

Pass the paint cap parameter through to text rendering so
that dashed text draws correctly.

R=bungeman@google.com
BUG=226341
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1776983002

Review URL: https://codereview.chromium.org/1776983002

Fix compile on Windows swarming bot

BUG=skia:4763
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1791513002

Review URL: https://codereview.chromium.org/1791513002

Add CMake builder to CQ

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1810523002

Review URL: https://codereview.chromium.org/1810523002

add missing API

SkMemoryWStream needs SK_API to be accessible from PDFium.

R=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1808683002

Review URL: https://codereview.chromium.org/1808683002

Create new buffer (if necessary) on vkMap

The assumption here is that we only read via a Map after we
sync the command buffer, and that when writing we don't Map
to update, but to overwrite. This appears to be true.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1774913005

Review URL: https://codereview.chromium.org/1774913005