optimize the disable-caching case for SkImage::readPixels

If the client wants no caching, and we haven't already cached it, pass the
caller's dst-buffer directly down to the generator, avoiding the (previous)
extra memcpy.

BUG=skia:4594

Review URL: https://codereview.chromium.org/1473373002

Whitespace change to trigger bots after gsutil change

TBR=rmistry
BUG=556703

Review URL: https://codereview.chromium.org/1471373002

Add sniff for rect, oval & rrect to SkCanvas::clipPath

Review URL: https://codereview.chromium.org/1466303004

Add offscreen mode to visualbench

Replaces "nvpr mode" with "offscreen mode" and a flag to enable/
disable nvpr. Assigns new config names to the various combinations of
flags and begins printing them with the bench results in order to match
nanobench.

BUG=skia:

Review URL: https://codereview.chromium.org/1473253002

Optimize visualbench offscreen blits

Modifies NvprWrappedBenchmark to present its offscreen canvas via
GrContext::copySurface rather than going through the SkCanvas API.
Only copies a subrectangle the size of the benchmark, not the entire
canvas.

BUG=skia:

Review URL: https://codereview.chromium.org/1443263002

Change XPS to use find and place glyph.

Testing:
The testing I did is non-standard. I just compared images using XPS viewer.
They all looked the same, but the XPS device is suffering from bit rot.

BUG=skia:

Review URL: https://codereview.chromium.org/1471043002

Sniff out rects, ovals & rrects in SkGpuDevice::drawPath

Review URL: https://codereview.chromium.org/1468423002

Generate list of GPU contexts outside ImageTest tests

Use DEF_GPUTEST_FOR_*_CONTEXT macros to obtain the
test GPU context.

Makes changing the context -related classes easier,
since not all tests need to be changed.

BUG=skia:2992

Review URL: https://codereview.chromium.org/1454663002

Cherry pick one more commit into libjpeg-turbo

I forgot to add the last fix for the skip scanlines patch.

BUG=skia:

Review URL: https://codereview.chromium.org/1476483002

Upgrade debugger to print more SkPath info (isOval & isRRect)

Review URL: https://codereview.chromium.org/1470063002

skia: Holding onto the image in the benchmark causes issues

Skia'a nanobench can hold onto an image, that causes the lifetime
of GrGLGpu to extend past that of the GL interface.

Kill reference to surface and image before killing the interface

Review URL: https://codereview.chromium.org/1472433002

Clarify when oval & rrects get devolved to paths in SkGpuDevice/GrDrawContext boundary

The GrDrawContext drawOval and drawRRect methods are only called from SkGpuDevice so there is no value in delaying the conversion to a path when dashing is involved.

Review URL: https://codereview.chromium.org/1470103002

Add SK_BEGIN_REQUIRE_DENSE / SK_END_REQUIRE_DENSE.

Use in on a few examples of things we hash.

BUG=skia:

Review URL: https://codereview.chromium.org/1468483002

Whitespace change to trigger android automerger

Review URL: https://codereview.chromium.org/1473233002 .

Appeasing the compilers

NOTRY=True
NOTREECHECKS=True

TBR=robertphillips@google.com

BUG=skia:

Review URL: https://codereview.chromium.org/1471143002

Fix for recent SrcOver XP Change

NOTREECHECKS=True
NOTRY=True

BUG=skia:

Review URL: https://codereview.chromium.org/1473443004

Make stencil and cover path rendering have more reasonable bounds.

BUG=skia:

Review URL: https://codereview.chromium.org/1471883002

Remove unecessary GetGlyphIdMetrics call

BUG=skia:

Review URL: https://codereview.chromium.org/1467133003

Move command line print above parse in VisualBench

TBR=bsalomon@google.com
BUG=skia:

Review URL: https://codereview.chromium.org/1467383003

Don't create a GXPFactory when blend is SrcOver

BUG=skia:

Review URL: https://codereview.chromium.org/1471053002

Move GrAtlasTextBatch blob regeneration to template

This will remove the runtime branches.  I don't think it makes the
code any harder to read either

BUG=skia:

Review URL: https://codereview.chromium.org/1466333003

scaling API on SkPixmap

BUG=skia:4481

Review URL: https://codereview.chromium.org/1463373002

Add stroking support to distance field path renderer

Also slightly increases sizes of paths accepted for distance field
caching.

Committed: https://skia.googlesource.com/skia/+/5ce76efd1c847308c7bcac17bd87d567c42cd786

Committed: https://skia.googlesource.com/skia/+/73ee6770260dbeeabc4a78eee4f13533f0031211

Review URL: https://codereview.chromium.org/1460873002

Restrict query bounds for reduce clip to dev bounds

Review URL: https://codereview.chromium.org/1467253002

Switch Skia's libjpeg-turbo to 1.4.2

The Android security team wants Android to be on the latest version
of libjpeg-turbo, so Skia should test on this version as well.

This shouldn't affect Gold.

I wouldn't be surprised if this causes a few ASAN/valgrind issues
that we need to fix.

BUG=skia:

Committed: https://skia.googlesource.com/skia/+/8e8eda9d4d2e6e65e3339eaaaf8144a4ef6c721c

Review URL: https://codereview.chromium.org/1469553002

use malloc/calloc/free instead of moz_malloc/moz_calloc/moz_free in ports/SkMemory_mozalloc.cpp

BUG=skia:4560

Review URL: https://codereview.chromium.org/1449093002

Revert of skstd -> std for unique_ptr (patchset #16 id:300001 of https://codereview.chromium.org/1436033003/ )

Reason for revert:
Perf-Mac10.9-Clang-MacMini6.2-CPU-AVX-x86_64-Release has not updated yet.

Original issue's description:
> skstd -> std for unique_ptr
>
> TBR=reed@google.com
> No public API changes.
>
> BUG=skia:4564
>
> Committed: https://skia.googlesource.com/skia/+/755c553c17b82bb5de3d9cc8d3b2a866ff9e9e50
>
> CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Mac10.9-Clang-x86_64-Release-CMake-Trybot,Build-Ubuntu-GCC-x86_64-Debug-CrOS_Link-Trybot
>
> Committed: https://skia.googlesource.com/skia/+/06189155d987db5c7e69015f6ea87c2168d6a065

TBR=bungeman@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4564

Review URL: https://codereview.chromium.org/1467333002

skstd -> std for unique_ptr

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

BUG=skia:4564

Committed: https://skia.googlesource.com/skia/+/755c553c17b82bb5de3d9cc8d3b2a866ff9e9e50

CQ_EXTRA_TRYBOTS=client.skia.compile:Build-Mac10.9-Clang-x86_64-Release-CMake-Trybot,Build-Ubuntu-GCC-x86_64-Debug-CrOS_Link-Trybot

Review URL: https://codereview.chromium.org/1436033003

Avoid devolving to a path when conservative clipping with RRects

Review URL: https://codereview.chromium.org/1461923004

Take span len into account when determining whether gradient is vertical

R=reed@google.com

Review URL: https://codereview.chromium.org/1471753004

Add SkPngChunkReader.

This class allows a client of SkCodec to read chunks in the data
stream that are not recognized by libpng. This is used by Android
to specify ninepatch data.

Taken from SkImageDecoder::Peeker. Modify the name of the class
and its method to be more specific to their use. Make
SkImageDecoder::Peeker a subclass of the new class, to help stage
the change in Android.

Add a test to verify that it works.

BUG=skia:4574
BUG=skia:3257

Committed: https://skia.googlesource.com/skia/+/3389e00136188800b98ca69488c0418c374fd78b

Review URL: https://codereview.chromium.org/1040453002

Revert of Add SkPngChunkReader. (patchset #9 id:160001 of https://codereview.chromium.org/1040453002/ )

Reason for revert:
Busted Chromium builds:

../../third_party/skia/src/ports/SkImageDecoder_empty.cpp:63:17: error: no type
named 'Peeker' in 'SkImageDecoder'
SkImageDecoder::Peeker* SkImageDecoder::setPeeker(Peeker*) {
~~~~~~~~~~~~~~~~^
../../third_party/skia/src/ports/SkImageDecoder_empty.cpp:63:51: error: unknown
type name 'Peeker'
SkImageDecoder::Peeker* SkImageDecoder::setPeeker(Peeker*) {

Original issue's description:
> Add SkPngChunkReader.
>
> This class allows a client of SkCodec to read chunks in the data
> stream that are not recognized by libpng. This is used by Android
> to specify ninepatch data.
>
> Taken from SkImageDecoder::Peeker. Modify the name of the class
> and its method to be more specific to their use. Make
> SkImageDecoder::Peeker a subclass of the new class, to help stage
> the change in Android.
>
> Add a test to verify that it works.
>
> BUG=skia:4574
> BUG=skia:3257
>
> Committed: https://skia.googlesource.com/skia/+/3389e00136188800b98ca69488c0418c374fd78b

TBR=djsollen@google.com,reed@google.com,msarett@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:4574

Review URL: https://codereview.chromium.org/1472863003

Revert of Switch Skia's libjpeg-turbo to 1.4.2 (patchset #2 id:160001 of https://codereview.chromium.org/1469553002/ )

Reason for revert:
Breaking jpeg decodes everywhere.  I need to look into this.

Original issue's description:
> Switch Skia's libjpeg-turbo to 1.4.2
>
> The Android security team wants Android to be on the latest version
> of libjpeg-turbo, so Skia should test on this version as well.
>
> This shouldn't affect Gold.
>
> I wouldn't be surprised if this causes a few ASAN/valgrind issues
> that we need to fix.
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/8e8eda9d4d2e6e65e3339eaaaf8144a4ef6c721c

TBR=scroggo@google.com,djsollen@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/1468073002

Add SkPngChunkReader.

This class allows a client of SkCodec to read chunks in the data
stream that are not recognized by libpng. This is used by Android
to specify ninepatch data.

Taken from SkImageDecoder::Peeker. Modify the name of the class
and its method to be more specific to their use. Make
SkImageDecoder::Peeker a subclass of the new class, to help stage
the change in Android.

Add a test to verify that it works.

BUG=skia:4574
BUG=skia:3257

Review URL: https://codereview.chromium.org/1040453002

Review URL: https://codereview.chromium.org/1464593004

Switch Skia's libjpeg-turbo to 1.4.2

The Android security team wants Android to be on the latest version
of libjpeg-turbo, so Skia should test on this version as well.

This shouldn't affect Gold.

I wouldn't be surprised if this causes a few ASAN/valgrind issues
that we need to fix.

BUG=skia:

Review URL: https://codereview.chromium.org/1469553002

Fix GLInterfaceValidation test after "Remove GrContextFactory::getGLContext"

Fix GLInterfaceValidation test crash after "Remove
GrContextFactory::getGLContext" commit.

Review URL: https://codereview.chromium.org/1464283002

Remove GrContextFactory::getGLContext

Remove GrContextFactory::getGLContext, it is problematic:

It has the bug of not checking for the context type.

It also is error-prone, since the GL context is not made
current, but it is callers may assume it is current.

It is also not used very much.

Clients can use GrContextFactory::getContextInfo.

BUG=skia:

Review URL: https://codereview.chromium.org/1455093003

Initial version of external_oes texture support and unit test

Committed: https://skia.googlesource.com/skia/+/27a048700778d4cebfc23301d1780649791b0e03

Review URL: https://codereview.chromium.org/1451683002

use pinned value during lerp for vertical gradients

slight formatting clean up on associated gm

BUG=skia:

Review URL: https://codereview.chromium.org/1471543002

SkPDF: add `final` keyword to leaf classes

Review URL: https://codereview.chromium.org/1461403002

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=
NO_MERGE_BUILDS

Review URL: https://codereview.chromium.org/1469693002

Revert of Add stroking support to distance field path renderer (patchset #8 id:160001 of https://codereview.chromium.org/1460873002/ )

Reason for revert:
Valgrind failures:

http://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-ShuttleA-GPU-GTX550Ti-x86_64-Release-Valgrind/builds/420/steps/dm/logs/stdio

==7062== Use of uninitialised value of size 8
==7062==    at 0x8ECA9A: AADistanceFieldPathBatch::onPrepareDraws(GrVertexBatch::Target*) (SkTDynamicHash.h:94)
==7062==    by 0x90DCF5: GrVertexBatch::onPrepare(GrBatchFlushState*) (GrVertexBatch.cpp:16)
==7062==    by 0x8B592A: GrDrawTarget::prepareBatches(GrBatchFlushState*) (GrBatch.h:100)
==7062==    by 0x8B51EA: GrDrawingManager::flush() (GrDrawingManager.cpp:78)
==7062==    by 0x8AE568: GrContext::flush(int) (GrContext.cpp:191)
==7062==    by 0x6C118A: GrContext::getTestTarget(GrTestTarget*, GrRenderTarget*) (GrTest.cpp:56)
==7062==    by 0x4F4D82: GrDrawingManager::ProgramUnitTest(GrContext*, int) (GLProgramsTest.cpp:360)
==7062==    by 0x4F58EC: test_GLPrograms(skiatest::Reporter*, GrContextFactory*) (GLProgramsTest.cpp:450)
==7062==    by 0x40925E: run_test(skiatest::Test*) (DM.cpp:1022)
==7062==    by 0x40BAC2: run_enclave_and_gpu_tests(SkTArray<Task, false>*) (DM.cpp:1033)
==7062==    by 0x7621A9: SkTaskGroup::wait() (SkTaskGroup.cpp:88)
==7062==    by 0x40B756: dm_main() (DM.cpp:1124)
==7062==    by 0x40BA5F: main (DM.cpp:1211)
==7062==  Uninitialised value was created by a stack allocation
==7062==    at 0x8DB484: GrTest::TestStrokeRec(SkRandom*) (GrTestUtils.cpp:232)
==7062==
{
   <insert_a_suppression_name_here>
   Memcheck:Value8
   fun:_ZN24AADistanceFieldPathBatch14onPrepareDrawsEPN13GrVertexBatch6TargetE
   fun:_ZN13GrVertexBatch9onPrepareEP17GrBatchFlushState
   fun:_ZN12GrDrawTarget14prepareBatchesEP17GrBatchFlushState
   fun:_ZN16GrDrawingManager5flushEv
   fun:_ZN9GrContext5flushEi
   fun:_ZN9GrContext13getTestTargetEP12GrTestTargetP14GrRenderTarget
   fun:_ZN16GrDrawingManager15ProgramUnitTestEP9GrContexti
   fun:_ZL15test_GLProgramsPN8skiatest8ReporterEP16GrContextFactory
   fun:_ZL8run_testPN8skiatest4TestE
   fun:_ZL25run_enclave_and_gpu_testsP8SkTArrayI4TaskLb0EE
   fun:_ZN11SkTaskGroup4waitEv
   fun:_Z7dm_mainv
   fun:main
}

Original issue's description:
> Add stroking support to distance field path renderer
>
> Also slightly increases sizes of paths accepted for distance field
> caching.
>
> Committed: https://skia.googlesource.com/skia/+/5ce76efd1c847308c7bcac17bd87d567c42cd786
>
> Committed: https://skia.googlesource.com/skia/+/73ee6770260dbeeabc4a78eee4f13533f0031211

TBR=robertphillips@google.com,jvanverth@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1465053002

Revert[2] of "quick-accept clipRect calls"

This reverts commit ac8cabd729c75fdf774c0701b6ea7fd3ea5b9382.

BUG=skia:
TBR=

Review URL: https://codereview.chromium.org/1470563003

Revert of Initial version of external_oes texture support and unit test (patchset #14 id:260001 of https://codereview.chromium.org/1451683002/ )

Reason for revert:
Error wrapping external texture in GrTexture on some devices.

Original issue's description:
> Initial version of external_oes texture support and unit test
>
> Committed: https://skia.googlesource.com/skia/+/27a048700778d4cebfc23301d1780649791b0e03

TBR=joshualitt@google.com,egdaniel@google.com,robertphillips@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1459323004

Initial version of external_oes texture support and unit test

Review URL: https://codereview.chromium.org/1451683002

Revert of option to quick-accept clipRect calls (patchset #3 id:40001 of https://codereview.chromium.org/1458913005/ )

Reason for revert:
Breaking tests on Android:

D/skia    ( 4469):
D/skia    ( 4469): Failures:
D/skia    ( 4469):  ../../../tests/CanvasStateTest.cpp:325 clipStackBounds.width() == WIDTH
D/skia    ( 4469):  ../../../tests/CanvasStateTest.cpp:326 clipStackBounds.height() == HEIGHT
D/skia    ( 4469): 2 failures

http://build.chromium.org/p/client.skia.android/builders/Test-Android-GCC-NexusPlayer-CPU-SSE4-x86-Debug/builds/1387/steps/dm/logs/stdio

Original issue's description:
> option to quick-accept clipRect calls
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/90ba095c459e38581353073826785074b5953b8c

TBR=robertphillips@google.com,fmalita@chromium.org,reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review URL: https://codereview.chromium.org/1471513002

Revert of Make specialized SrcOver XPFactory (patchset #3 id:40001 of https://codereview.chromium.org/1455273006/ )

Reason for revert:
breaking some builds

Original issue's description:
> Make specialized SrcOver XPFactory
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/a7006d45217d128a94fa53cb4b827cca79bc7049

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

Review URL: https://codereview.chromium.org/1464933003

fix pathops coincidence fuzz bug

Simplifying a series of rects with very large bounds
triggers a coincidence bug where, after one of the
intersection points that marks a coincident range
has been deleted, it is referenced.

Both the deletion and reference is (probably) happening
in the SkOpCoincidence::AddExpanded() phase of
HandleCoincidence(), and may signify a bug that could
happen with usable input data, but I haven't been
able to determine that.

For now, abort the Simplify() when the erroneous
condition is detected.

TBR=reed@google.com
BUG=558281

Review URL: https://codereview.chromium.org/1463923002

Add stroking support to distance field path renderer

Also slightly increases sizes of paths accepted for distance field
caching.

Committed: https://skia.googlesource.com/skia/+/5ce76efd1c847308c7bcac17bd87d567c42cd786

Review URL: https://codereview.chromium.org/1460873002

Create GrGLSLVaryingHandler class for program building

BUG=skia:

Review URL: https://codereview.chromium.org/1462123003

Make specialized SrcOver XPFactory

BUG=skia:

Review URL: https://codereview.chromium.org/1455273006

Tiny linear gradient GM

A GM to expose issues with vertical linear gradients when points
and/or positions are too close.

R=reed@google.com

Review URL: https://codereview.chromium.org/1464693002

Fix UB in SkDivBits

This used to:
  DIVBITS_ITER was shifting bits up into the sign bit, which is a no-no.
  This turns numer into a uint32_t to make those defined, and adds a few notes.

  x >= 0 is always true for unsigned x, so we needed a few small logic refactors.

Instead it now:
  Only call SkDivBits if the old behavior is required.
  Usually, just do the divide with /.

BUG=skia:3562

Committed: https://skia.googlesource.com/skia/+/988adddd48322bfa3e3cb0c017cfce71fbbf1123

Committed: https://skia.googlesource.com/skia/+/6c7b104b4c08ae2332a6ce3c8c906da4e8c51e5f

TBR=reed@google.com
No API change.

Review URL: https://codereview.chromium.org/1455163004

Add SkNx_cast().

SkNx_cast() can cast between any of our vector types,
provided they have the same number of elements.

Any types should work with the default implementation,
and we can drop in specializations as needed, like the
SSE and NEON Sk4f -> Sk4i I included here as an example.

To make this work, I made some internal name changes:
    SkNi<N,T> -> SkNx<N, T>
    SkNf<N>   -> SkNx<N, float>
User aliases (Sk4f, Sk16b, etc.) stay the same.
We can land this first (it's PS1) if that makes things easier.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1464623002

Move glyph choosing to the find and place glyph code.

This duplicates the functionality of the (private)
SkPaint::getDrawCacheProc method into
SkFindAndPlaceGlyph::LookupGlyph. Eventually LookupGlyph
should replace getDrawCacheProc, at which point it should be
removed. The remaining users are gpu and pdf.

Review URL: https://codereview.chromium.org/1458193003

SkFontConfigInterface::createTypeface()

BUG=524578
R=bungeman@google.com,reed@google.com

Review URL: https://codereview.chromium.org/1454343003

Eliminate SkFILE: it always is the same as FILE.

Review URL: https://codereview.chromium.org/1467533003

option to quick-accept clipRect calls

BUG=skia:

Review URL: https://codereview.chromium.org/1458913005

enable conservative raster clipping

There may be other (better?) ways to enable this. For example, if we queried the device *before* we init our root DeviceCM, we wouldn't need this flag in InitFlags, since the device already has a virtual to opt-in ...

BUG=skia:

Review URL: https://codereview.chromium.org/1458153003

Initialize memory for BitMask and Clip test.

BUG=skia:

Review URL: https://codereview.chromium.org/1466713002

Factor out GrAtlasTextBatch fromt GrAtlasTextContext

BUG=skia:

Review URL: https://codereview.chromium.org/1458233003

Pack and align SkStrokeRec to 4-byte boundary.

The new key for the distance field path cache will contain an
SkStrokeRec. This change guarantees that we don't have any hidden
padding that has garbage values, thereby preventing apparently
equal keys from hashing to two different values. This also has
the nice effect of reducing the size of SkStrokeRec from 24 bytes
to 16 bytes.

Review URL: https://codereview.chromium.org/1465773003

Generate list of GPU contexts outside SurfaceTest tests

Add support for feeding the tests with contexts directly to the unit
test framework.

This fixes the problem where tests are more complex than needed just in
order to run the test code with multiple backends.

Also makes it possible to change the logic how contexts are
created. Instead of direct numbering, the different testable contexts
may be generated from filtered cross-product of context options. For
example: currently NVPR is a type of context. However, it could be also
an on/off feature of any context. In order to test this kind of context,
the enumeration can not be just of context type. It's simpler
to move the enumeration out of the tests.

A test targeting both normal and GPU backends would look like:

static void test_obj_behavior(skiatest::Reporter* reporter,
    SkObj* obj, [other params] ) {
    ... test with obj and param ..
}
DEF_TEST(ObjBehavior, reporter) {
    for (auto& object : generate_object) {
        for (auto& other_param : generate_other_variant) {
   test_obj_behavior(reporter, object, other_param);
}
    }
}
#if SK_SUPPORT_GPU
DEF_GPUTEST_FOR_ALL_CONTEXTS(ObjBehavior_Gpu, reporter, context) {
    for (auto& object : generate_gpu_object) {
        for (auto& other_param : generate_other_variant) {
   test_obj_behavior(reporter, object, other_param);
}
    }
}
#endif

Uses the feature in SurfaceTests as an example.

Moves SkSurface -related tests from ImageTest to SurfaceTest.

BUG=skia:2992

Review URL: https://codereview.chromium.org/1446453003

Revert of Fix UB in SkDivBits (patchset #4 id:50001 of https://codereview.chromium.org/1455163004/ )

Reason for revert:
This likely causes unit tests to break in depsroll.

See error:

http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/137723/steps/cc_unittests%20%28with%20patch%29/logs/stdio

 RUN      ] ImageScaledRenderSurface.ImageRenderSurfaceScaled_Software

[4616:3480:1119/165044:41067403:ERROR:pixel_comparator.cc(175)] Percentage of pixels with an error: 11.3967

[4616:3480:1119/165044:41067403:ERROR:pixel_comparator.cc(177)] Percentage of pixels with errors not greater than 0: 0

[4616:3480:1119/165044:41067403:ERROR:pixel_comparator.cc(180)] Average absolute error (excluding identical pixels): R=0.0600565 G=238.962 B=238.934 A=0

[4616:3480:1119/165044:41067403:ERROR:pixel_comparator.cc(185)] Largest absolute error: R=1 G=255 B=255 A=0

[4616:3480:1119/165044:41067403:ERROR:pixel_comparator.cc(203)] Error Bounding Box : 47,47 206x206

[4616:3480:1119/165044:41067419:ERROR:pixel_test_utils.cc(79)] Pixels do not match!

[4616:3480:1119/165044:41067419:ERROR:pixel_test_utils.cc(80)] Actual: data:image/png;base64,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

[4616:3480:1119/165044:41067419:ERROR:pixel_test_utils.cc(81)] Expected: data:image/png;base64,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

e:\b\build\slave\win\build\src\cc\test\layer_tree_pixel_test.cc(116): error: Value of: MatchesPNGFile(*result_bitmap_, ref_file_path, *pixel_comparator_)

  Actual: false

Expected: true

[  FAILED  ] ImageScaledRenderSurface.ImageRenderSurfaceScaled_Software (57 ms)

Original issue's description:
> Fix UB in SkDivBits
>
> DIVBITS_ITER was shifting bits up into the sign bit, which is a no-no.
> This turns numer into a uint32_t to make those defined, and adds a few notes.
>
> x >= 0 is always true for unsigned x, so we needed a few small logic refactors.
>
> BUG=skia:3562
>
> Committed: https://skia.googlesource.com/skia/+/988adddd48322bfa3e3cb0c017cfce71fbbf1123
>
> Committed: https://skia.googlesource.com/skia/+/6c7b104b4c08ae2332a6ce3c8c906da4e8c51e5f

TBR=caryclark@google.com,mtklein@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:3562

Review URL: https://codereview.chromium.org/1467513002

Implement SkGifCodec::onSkipScanlines()

This should give a performance improvment because we are able to skip
swizzling pixels.

This should also fix valgrind failures caused by color table lookups
with uninitialized memory.

BUG=skia:4270

Review URL: https://codereview.chromium.org/1460073002

add SkPath::isRRect

Add helper to track when a round rect was added to a path,
and then return the SkRRect specification that describes it.

Move the implementation for SkPath::RawIter to SkPathRef so it can be used there as well.

R=reed@google.com,robertphillips@google.com

Review URL: https://codereview.chromium.org/1461763004

Avoid geometric ninepatch if rendertarget has MSAA

BUG=skia:4586

Review URL: https://codereview.chromium.org/1458883009

Add more batch information printouts

Review URL: https://codereview.chromium.org/1457373003

Revert of Add stroking support to distance field path renderer (patchset #7 id:120001 of https://codereview.chromium.org/1460873002/ )

Reason for revert:
Asserting in nanobench and dm.

Original issue's description:
> Add stroking support to distance field path renderer
>
> Also slightly increases sizes of paths accepted for distance field
> caching.
>
> Committed: https://skia.googlesource.com/skia/+/5ce76efd1c847308c7bcac17bd87d567c42cd786

TBR=robertphillips@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1458233002

Add stroking support to distance field path renderer

Also slightly increases sizes of paths accepted for distance field
caching.

Review URL: https://codereview.chromium.org/1460873002

Record concat as Concat.

This seems to cause very slight diffs on GMs.
https://gold.skia.org/search2?issue=1462983002&unt=true&query=source_type%3Dgm&master=false

I'm not sure I understand why.
The diffs to at least pictureshader look like improvements.

BUG=skia:4584

Review URL: https://codereview.chromium.org/1462983002

Fix nearly-vertical gradient assert

Use a SkScalarNearlyZero(dx) test instead of !SkScalarIsFinite(invDx).

R=reed@google.com

Review URL: https://codereview.chromium.org/1456783005

add SkRecordOptimize2 and an experimental API for more aggressive opts

BUG=skia:

Review URL: https://codereview.chromium.org/1462973002

Remove unsigned < signed comparison.

This triggers -Wsign-compare on some builds.
Not sure why none of those builds have -Werror set.

BUG=skia:4561

Review URL: https://codereview.chromium.org/1464543002

Parametric contour start GM

A GM to capture the newly added SkPath API.

BUG=chromium:315277
R=caryclark@google.com,reed@google.com

Review URL: https://codereview.chromium.org/1457503002

Fix UB in SkDivBits

DIVBITS_ITER was shifting bits up into the sign bit, which is a no-no.
This turns numer into a uint32_t to make those defined, and adds a few notes.

x >= 0 is always true for unsigned x, so we needed a few small logic refactors.

BUG=skia:3562

Committed: https://skia.googlesource.com/skia/+/988adddd48322bfa3e3cb0c017cfce71fbbf1123

Review URL: https://codereview.chromium.org/1455163004

Fix GpuDrawPathTest on ANGLE and enable it

Fix GpuDrawPathTest on ANGLE by avoiding a crash
if MSAA surface creation fails.

Enable the test.

BUG=skia:4581

Review URL: https://codereview.chromium.org/1463493002

Revert float xfermodes back to Sk4f (from Sk8f).

Generally this was a performance win, even on devices without AVX due
to unrolling, but on ARM+NEON it looks like that unrolling hurt a bit.

    while (...) { blend a pixel }
~~~>
    while (...) { blend two pixels }
    if (n % 2) { blend last pixel }

BUG=chromium:555278
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1465483002

Fix BMP RLE bug: attempt 2

BUG=skia:

Review URL: https://codereview.chromium.org/1457213003

Modernize SkRecordPattern.h

  - Fold Or, Or3, Or4 into one flexible Or.
  - Fold Pattern1...Pattern7 into one flexible Pattern.
  - Rename Star Greedy

Still fighting with a flexible get<N>() method instead of first, second, third, etc.

BUG=skia:

Review URL: https://codereview.chromium.org/1465443002

Fix maybe unintialized warning for GrAutoLocaleSetter

In function void test_GLPrograms(skiatest::Reporter*, GrContextFactory*):
../../../src/gpu/GrAutoLocaleSetter.h:47:35: error:
als.GrAutoLocaleSetter::fOldLocale may be used uninitialized in this
function [-Werror=maybe-uninitialized]

From Build-Ubuntu-GCC-x86_64-Release-Trybot

Review URL: https://codereview.chromium.org/1456383002

add --write option to dump_record

BUG=skia:

Review URL: https://codereview.chromium.org/1459823003

Change prealloc number of child FPs be 1 rather than 2

TBR=joshualitt@google.com

Review URL: https://codereview.chromium.org/1458913003 .

Preallocate room for 2 child processors in FPs

Review URL: https://codereview.chromium.org/1456263003

Shift up bottom path in smallpaths GM.

This stops it from being cut off in the image.

Review URL: https://codereview.chromium.org/1462903002

Unit test for conic weight GrPath key collision

BUG=skia:4580
R=robertphillips@google.com,bsalomon@google.com

Review URL: https://codereview.chromium.org/1454223002

Add SkRecord::defrag().

Called by SkRecordOptimize(), this moves all the NoOps to the end and
slices them off.

This implementation with std::remove_if() is linear and doesn't malloc.

No diffs: https://gold.skia.org/search2?issue=1461663003&unt=true&query=source_type%3Dgm&master=false

BUG=skia:

Review URL: https://codereview.chromium.org/1461663003

Move glsl onto EmitArgs struct for emitCode

BUG=skia:

Review URL: https://codereview.chromium.org/1453623003

Fix leak in drawBitmapNine

NOTREECHECKS=true

BUG=skia:

Review URL: https://codereview.chromium.org/1462893002

[Reland] Fix NVPR assert for equivalent ovals

For oval paths, GrPath ignores the point order and only uses the bounds
when building its key.  This is problematic because

1) point order is important when dashing
2) GrStencilAndCoverPathRenderer asserts that the lookup SkPath is equal
   to the cached SkPath - which is not the case for ovals with different
   directions/different point order.

With this CL we no longer use the reduced oval key when dashing, and
instead fall through to the more general path cases.  The assert is
adjusted to accommodate "equivalent" ovals (when not dashing).

Also re-enabled & updated the GpuDrawPath unit test (disabled in
https://codereview.chromium.org/1456463003/, presumably due to the use
of uninitialized SkRects).

R=bsalomon@google.com,robertphillips@google.com,cdalton@nvidia.com

Review URL: https://codereview.chromium.org/1457073002

Make SkTLList prealloc its first block of nodes

Review URL: https://codereview.chromium.org/1458703005

Make block size a template parameter of SkTLList

Review URL: https://codereview.chromium.org/1457123002

Revert SkBlitMask_opts.h back to hand-coded NEON.

SkPx has triggered a bunch of small (2-9%) regressions on NEON devices.

BUG=skia:
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1462783002

Revert of Fix NVPR assert for equivalent ovals (patchset #1 id:1 of https://codereview.chromium.org/1457073002/ )

Reason for revert:
Causes failures on Android and Win8:

...
( 137/1245MB      9) 73.9ms unit test  GpuLayerCachec:\0\build\slave\workdir\build\skia\include\private\skuniqueptr.h:164: failed assertion "get() != pointer()"

Caught exception 2147483651 EXCEPTION_BREAKPOINT
...

Original issue's description:
> Fix NVPR assert for equivalent ovals
>
> For oval paths, GrPath ignores the point order and only uses the bounds
> when building its key.  This is problematic because
>
> 1) point order is important when dashing
> 2) GrStencilAndCoverPathRenderer asserts that the lookup SkPath is equal
>    to the cached SkPath - which is not the case for ovals with different
>    directions/different point order.
>
> With this CL we no longer use the reduced oval key when dashing, and
> instead fall through to the more general path cases.  The assert is
> adjusted to accommodate "equivalent" ovals (when not dashing).
>
> Also re-enabled & updated the GpuDrawPath unit test (disabled in
> https://codereview.chromium.org/1456463003/, presumably due to the use
> of uninitialized SkRects).
>
> R=bsalomon@google.com,robertphillips@google.com,cdalton@nvidia.com
>
> Committed: https://skia.googlesource.com/skia/+/f9b1577d763988ebc043ddabf80674f71571ecff

TBR=bsalomon@google.com,cdalton@nvidia.com,robertphillips@google.com,fmalita@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review URL: https://codereview.chromium.org/1461913002

simplify insertion methods for SkTLList

Review URL: https://codereview.chromium.org/1459663002

Fix NVPR assert for equivalent ovals

For oval paths, GrPath ignores the point order and only uses the bounds
when building its key.  This is problematic because

1) point order is important when dashing
2) GrStencilAndCoverPathRenderer asserts that the lookup SkPath is equal
   to the cached SkPath - which is not the case for ovals with different
   directions/different point order.

With this CL we no longer use the reduced oval key when dashing, and
instead fall through to the more general path cases.  The assert is
adjusted to accommodate "equivalent" ovals (when not dashing).

Also re-enabled & updated the GpuDrawPath unit test (disabled in
https://codereview.chromium.org/1456463003/, presumably due to the use
of uninitialized SkRects).

R=bsalomon@google.com,robertphillips@google.com,cdalton@nvidia.com

Review URL: https://codereview.chromium.org/1457073002

Revert of Fix UB in SkDivBits (patchset #2 id:10002 of https://codereview.chromium.org/1455163004/ )

Reason for revert:
Need to reland with #define guards for tiny layout test changes. (Yikes!)

Original issue's description:
> Fix UB in SkDivBits
>
> DIVBITS_ITER was shifting bits up into the sign bit, which is a no-no.
> This turns numer into a uint32_t to make those defined, and adds a few notes.
>
> x >= 0 is always true for unsigned x, so we needed a few small logic refactors.
>
> BUG=skia:3562
>
> Committed: https://skia.googlesource.com/skia/+/988adddd48322bfa3e3cb0c017cfce71fbbf1123

TBR=caryclark@google.com,mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:3562

Review URL: https://codereview.chromium.org/1457863002

In BUILD.public, remove compile-time dependency on GL.

BUG=skia:

Review URL: https://codereview.chromium.org/1458603006

Add small paths GM

Used to testing caching and quality of distance field paths.

Review URL: https://codereview.chromium.org/1456823003