Reduce overhead for linear color xforms

We used to build src and dst transfer fn tables
every time a new xform was created with linear
src and dst.  Now we don't compute them because
we don't need them.

This will make SkColorSpaceXform a far better
option for any xforms with float or half-float
inputs or outputs, particularly on a small number
of pixels.

This CL also moves SkColorSpaceXform closer to
what I anticipate will be the eventual 'API design'.
I think apply() will want to take a SrcColorType enum
(not created yet because it's not necessary yet) and
a DstColorType enum (still using SkColorType because
there's not yet a reason not to).

Performance changes:
toSRGB  341us -> 366us
to2Dot2 404us -> 403us
toF16   318us -> 304us

There's no reason for toSRGB or to2Dot2 to change.
The refactor seems to have caused the compiler to
order the instructions a little differently...
This is something to come back to if we need to
squeeze more performance out of sRGB.  For now,
let's not be held up by something we don't control.

F16 likely improves because we are no longer
(unnecessarily) building the linear tables.

Code size gets a little bigger.  Measuring
SkColorSpaceXform size as a percentage of src/ size,
we go from 0.8% to 1.4%.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2335723002

Review-Url: https://codereview.chromium.org/2335723002

Switch default for SkGaussianBlurShader radius size.

One step towards removing the 6.2 radius entirely.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2333403002
TBR=bsalomon@google.com

Review-Url: https://codereview.chromium.org/2333403002

GN: Android Vulkan support

Android's the easiest place to start supporting Vulkan, so it's up first.

ndk_api becomes user-specifiable so that the Vulkan bots can kick it up
to 24, the first release supporting Vulkan.  The defaults remain the same:
18 for 32-bit and 21 for 64-bit, the first release supporting 64-bit.

To test this, I set ndk_api=24 and skia_use_vulkan=true in GN, then
   $ ninja -C out dm; and droid out/dm --config vk --src gm

Seems to work!  Bot scripts to follow.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336343002

Review-Url: https://codereview.chromium.org/2336343002

update derive_compile_bot_name() for GN_Android_Vulkan

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336233005

Review-Url: https://codereview.chromium.org/2336233005

Android GN scripts: distingush directories for data and binaries.

This enables data reuse, e.g. https://luci-milo.appspot.com/swarming/task/313eff129c753d10 or https://luci-milo.appspot.com/swarming/task/313f17a174c6c710

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2335203004

Review-Url: https://codereview.chromium.org/2335203004

change write-image to use pipeverbs like everyone else, and add test

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2335973003

Review-Url: https://codereview.chromium.org/2335973003

re-add SkDebug_stdio for Fuchsia builds

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2331383004

Review-Url: https://codereview.chromium.org/2331383004

Simulated fonts aren't TrueType fonts.

Some font back-ends provide simulated fonts such as fake bold or fake
oblique. These fonts should not be reported as TrueType, since the font
data isn't what is actually used to draw the glyphs.

BUG=chromium:639198
BUG=chromium:614612
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2333423002

Review-Url: https://codereview.chromium.org/2333423002

[SVGDom] Linear gradient 'gradientTransform' support

R=stephana@google.com,robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2339533002

Review-Url: https://codereview.chromium.org/2339533002

Update gn_android_flavor for Vulkan.

C.f. https://codereview.chromium.org/2336343002 for basic GN support.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2333293004

Review-Url: https://codereview.chromium.org/2333293004

No \n in bench names.

BUG=skia:
R=mtklein
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2332153003

Review-Url: https://codereview.chromium.org/2332153003

Android GN: don't reboot before running

We're already rebooting after running, so we should already be ready to go when the task starts.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2334223003

Review-Url: https://codereview.chromium.org/2334223003

[SVGDom] Linear gradient 'spreadMethod' support

R=stephana@google.com,robertphillips@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2337203002

Review-Url: https://codereview.chromium.org/2337203002

Remove clear /data/local/tmp steps.

These were intended to prevent GN and GYP Android bots from stomping on
each other.  Turns out, they don't, even without this... they're writing
most files to completely separate paths:
  - GYP puts most data under $EXTERNAL_STORAGE, generally /sdcard
    and its binaries (libdm.so, libnanobench.so, libskia.so, skia_launcher)
    in /data/local/tmp;
  - GN puts everything under /data/local/tmp, and its binaries (dm, nanobench)
    don't overlap GYP's.

So clearing /data/local/tmp was essentially just removing GN's data; GYP's
data file caching was never affected and can't conflict with GN's.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2340473002

Review-Url: https://codereview.chromium.org/2340473002

Add ContextOptions field to SkCommandLineConfigGpu

Reduces duplicate code by computing the ContextOptions in one spot.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2337163002

Review-Url: https://codereview.chromium.org/2337163002

Use sk_sp text blob APIs

SkTextBlobBuilder::build()      -> make()
SkAutoTUnref<const SkTextBlob>  -> sk_sp<SkTextBlob>
drawTextBlob(const SkTextBlob*) -> drawTextBlob(const sk_sp<SkTextBlob>&)

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2335493005

Review-Url: https://codereview.chromium.org/2335493005

Lots of little cleanup improvements to Vulkan

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336763002

Review-Url: https://codereview.chromium.org/2336763002

Implement Fill() for incomplete decodes to RGBA_F16

Before this patch, we would hit an SkASSERT(false).

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2335203002

Review-Url: https://codereview.chromium.org/2335203002

Add Nexus5 to CQ as an experiment

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2332353002

Review-Url: https://codereview.chromium.org/2332353002

Add tests for gamma correction of gradient stops to gamma GM

Includes three different cases to exercise different cases in Ganesh.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336973003

Review-Url: https://codereview.chromium.org/2336973003

add pipecanvas

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2201323003

Review-Url: https://codereview.chromium.org/2201323003

SkPDF: SkTypeface_FreeType ToUnicode table improvement.

Currently the SkTypeface_FreeType::onGetAdvancedTypefaceMetrics
synthesized glyph to Unicode mapping returns the Unicode point
of the last character to map to the glyph. In practice it is
better to guess the first character to map to the glyph instead.

BUG=359065
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2329953003

Review-Url: https://codereview.chromium.org/2329953003

I missed an infra_step, pushing directories.

NOTREECHECKS=true

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2337793003

Review-Url: https://codereview.chromium.org/2337793003

Encode kIndex8 to PNG more efficiently

Only write non-opaque alpha values to the tRNS chunk.

BUG=skia:5616
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2330053002

Review-Url: https://codereview.chromium.org/2330053002

Mark infra steps as infra_step in GN recipe flavors.

NOTREECHECKS=true

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2329943004

Review-Url: https://codereview.chromium.org/2329943004

Fixup SkRRectsGaussianEdgeShader's shaders

This fixes some visual artifacts in the original CL.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2335783003

Review-Url: https://codereview.chromium.org/2335783003

[SVGDom] Initial linear gradient support

Kind of a big change, to connect several new bits into something useful:

  * ID tracking & lookup
  * new asPaint() node virtual to support shader (and in the future filter) based paint servers
  * <defs>, <linearGradient> and <stop> element support
  * 'href', 'offset', 'stop-color', 'stop-opacity' attribute support
  * IRI/FuncIRI and rgb(...) parsing

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2327233003

Review-Url: https://codereview.chromium.org/2327233003

Change shared_sources.gni to use relative path

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2333923003

Review-Url: https://codereview.chromium.org/2333923003

Disable png encodes from Alpha8, Float16

These don't behave as we would want anyway.  They just copy
to N32, and then encode.

BUG=skia:5616
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2332743003

Review-Url: https://codereview.chromium.org/2332743003

Alternate approach to turn on freetype in Fuchsia.

What do you think of this?  It's basically what you wrote, factored differently.

Flipping on skia_use_freetype will enable the SkFontHost_Freetype*.cpp files
via the :typeface_freetype target.

Then we just pull out the SkFontMgr Fuchsia wants to use into its own little
:fontmgr_fuchsia target.  No major difference here, except the availability
of freetype headers, etc, is a bit more tightly scoped to just those that
need them.

We (obviously?) don't have Fuchsia builders, so I've got to rely on you for sanity testing this.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2333913002

Review-Url: https://codereview.chromium.org/2333913002

Support RGBA/BGRA Premul/Unpremul from SkPNGImageEncoder

BUG=skia:5616
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2325223002

Review-Url: https://codereview.chromium.org/2325223002

Add back the GalaxyS3 and GalaxyS4 entries in recipes

S3 is needed for some recipe simulation stuff and in general having
these present does not hurt us.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336513003

Review-Url: https://codereview.chromium.org/2336513003

Remove unused GrGLSLGeometryProcessor::emitTransform variation
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2338453002

Review-Url: https://codereview.chromium.org/2338453002

Revert of SkLiteRecorder: don't tell SkCanvas about clips (patchset #1 id:1 of https://codereview.chromium.org/2227833004/ )

Reason for revert:
This CL breaks Android CTS tests CanvasStateTests#testClipPathReturnValues, CanvasStateTests#testClipRectReturnValues, and CanvasStateTests#testQuickReject.

Reverting it won't significantly affect Android's performance. See the following test result of testInvalidateTree Jank test. (Derek said that this is the only test closely related to this CL.)

**After reverting**
INSTRUMENTATION_STATUS: gfx-max-slow-draw=0.0
INSTRUMENTATION_STATUS: gfx-max-high-input-latency=0.0
INSTRUMENTATION_STATUS: gfx-max-slow-ui-thread=100.0
INSTRUMENTATION_STATUS: gfx-avg-slow-bitmap-uploads=100.0
INSTRUMENTATION_STATUS: gfx-avg-missed-vsync=72.96467105972435
INSTRUMENTATION_STATUS: gfx-avg-high-input-latency=0.0
INSTRUMENTATION_STATUS: gfx-max-slow-bitmap-uploads=100.0
INSTRUMENTATION_STATUS: gfx-max-missed-vsync=73.56181150550796
INSTRUMENTATION_STATUS: gfx-avg-frame-time-90=69.0
INSTRUMENTATION_STATUS: gfx-avg-frame-time-95=69.0
INSTRUMENTATION_STATUS: gfx-avg-frame-time-99=69.0
INSTRUMENTATION_STATUS: gfx-max-jank=100.0
INSTRUMENTATION_STATUS: gfx-avg-slow-draw=0.0
INSTRUMENTATION_STATUS: gfx-avg-slow-ui-thread=100.0
INSTRUMENTATION_STATUS: gfx-max-frame-time-90=69
INSTRUMENTATION_STATUS: gfx-max-frame-time-95=69
INSTRUMENTATION_STATUS: gfx-max-frame-time-99=69
INSTRUMENTATION_STATUS: gfx-avg-jank=100.0
INSTRUMENTATION_STATUS_CODE: -1
.
Test results for InstrumentationTestRunner=.
Time: 77.404

**Before Reverting**
gfx-max-slow-draw=1.1135857461024499
INSTRUMENTATION_STATUS: gfx-max-high-input-latency=0.0
INSTRUMENTATION_STATUS: gfx-max-slow-ui-thread=100.0
INSTRUMENTATION_STATUS: gfx-avg-slow-bitmap-uploads=100.0
INSTRUMENTATION_STATUS: gfx-avg-missed-vsync=70.79687591131983
INSTRUMENTATION_STATUS: gfx-avg-high-input-latency=0.0
INSTRUMENTATION_STATUS: gfx-max-slow-bitmap-uploads=100.0
INSTRUMENTATION_STATUS: gfx-max-missed-vsync=76.94174757281553
INSTRUMENTATION_STATUS: gfx-avg-frame-time-90=67.66666666666667
INSTRUMENTATION_STATUS: gfx-avg-frame-time-95=67.66666666666667
INSTRUMENTATION_STATUS: gfx-avg-frame-time-99=69.0
INSTRUMENTATION_STATUS: gfx-max-jank=100.0
INSTRUMENTATION_STATUS: gfx-avg-slow-draw=0.5317509395832639
INSTRUMENTATION_STATUS: gfx-avg-slow-ui-thread=100.0
INSTRUMENTATION_STATUS: gfx-max-frame-time-90=69
INSTRUMENTATION_STATUS: gfx-max-frame-time-95=69
INSTRUMENTATION_STATUS: gfx-max-frame-time-99=69
INSTRUMENTATION_STATUS: gfx-avg-jank=100.0
INSTRUMENTATION_STATUS_CODE: -1
.
Test results for InstrumentationTestRunner=.
Time: 77.411

OK (1 test)

Original issue's description:
> SkLiteRecorder: don't tell SkCanvas about clips
>
> If no one reads our clip, and we don't expect quickReject() to help,
> we can probably get away without maintaining a proper clip stack.
>
> This puts us at about 0.6-0.7x of previous record cost.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2227833004
>
> Committed: https://skia.googlesource.com/skia/+/8369e32a05543c2619b1d7e740fe6ff6d2f03af3

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

Review-Url: https://codereview.chromium.org/2335773002

Remove base class for installed transforms in GrPrimitiveProcessor
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336723002

Review-Url: https://codereview.chromium.org/2336723002

GN: Android perf/tests

Just to take inventory of which old problems still linger and which are now moot, I've gone out of my way to have this stand alone.  All of gn_android_flavor's logic is self contained, without any dependency on the platform_tools scripts.

The tricky bits turn out to be, copying directories containing symlinks---or really any copying involving more than one file---and getting the exit code back from adb shell.  Luckily the ADB I've got on my desktop and my Nexus 5x seems to handle this all without the awkward workarounds you see here, so there's hope that One Day Soon the weird parts (basically, anything with inline python) can go away.  Once we've got these bots landed green, I'll go see whether the fixes are due to ADB updates, Android updates, or perhaps something else like hardware.

The parts marked TEMPORARY are a nod to the fact that the devices are used by gn_android_flavor and android_flavor both today.  It's mostly about not stepping on each other's toes or leaving anything laying around that might confuse each other.  The marked parts can go away when bots are either gn_ or non-gn_ but not both.

I have omitted a few steps that may be important, but which are easy independent follow-ups:
  - running as root
  - locking clocks
  - waiting on battery levels
  - fancier wait-for-ready than adb wait-for-usb-device
It'd be nice to, e.g., reaffirm that locking clocks helps perf stability, and that we're locking to the best policy.  I've tried to keep this CL as trim as possible, leaving any of these vaguely optional steps for later.

As of PS 41 or so, it looks like the trybots are all behaving as expected.
We should expect no new images in Gold.  Can we see trybots in Perf yet?

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2320153002

CQ_INCLUDE_TRYBOTS=master.client.skia.android:Perf-Android-Clang-AndroidOne-CPU-MT6582-arm-Debug-GN_Android-Trybot,Perf-Android-Clang-AndroidOne-CPU-MT6582-arm-Release-GN_Android-Trybot,Perf-Android-Clang-AndroidOne-GPU-Mali400MP2-arm-Debug-GN_Android-Trybot,Perf-Android-Clang-AndroidOne-GPU-Mali400MP2-arm-Release-GN_Android-Trybot,Test-Android-Clang-AndroidOne-CPU-MT6582-arm-Debug-GN_Android-Trybot,Test-Android-Clang-AndroidOne-CPU-MT6582-arm-Release-GN_Android-Trybot,Test-Android-Clang-AndroidOne-GPU-Mali400MP2-arm-Debug-GN_Android-Trybot,Test-Android-Clang-AndroidOne-GPU-Mali400MP2-arm-Release-GN_Android-Trybot

Review-Url: https://codereview.chromium.org/2320153002

Gamut transformation of the paint color in Ganesh

Conversion from sRGB to destination gamut is going to be very common,
so I'm caching that xform (if there is one) on the draw context.

Results verified in the gamut GM (two more boxes correct).

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2330553003

Review-Url: https://codereview.chromium.org/2330553003

Match Android framework's non-AA point and line offset

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336623002

Review-Url: https://codereview.chromium.org/2336623002

Add shader-based GaussianEdgeShader for reveal case

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2321713004

Review-Url: https://codereview.chromium.org/2321713004

Remove GrPrimitiveProcessor::hasTransformedLocalCoords
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2327613002

Review-Url: https://codereview.chromium.org/2327613002

change SkStreams to work with sk_sp<SkData> instead of SkData*

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2333713002

Review-Url: https://codereview.chromium.org/2333713002

Fix device cfg for GalaxyS7

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2330093002

Review-Url: https://codereview.chromium.org/2330093002

Documentation: MSKP

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2332783002
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=2332783002

Review-Url: https://codereview.chromium.org/2332783002

Add workaround for Adreno copyImageToBuffer calls on Vulkan

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2328823003

Review-Url: https://codereview.chromium.org/2328823003

Run recipes.py with --timestamps

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2330123002

Review-Url: https://codereview.chromium.org/2330123002

Fix GaussianEdgeShader GLSL code

Implicit conversion of int constants to float is not always
guaranteed

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2328263004
TBR=herbderby@google.com
NOTRY=true
NOTREECHECKS=true

Review-Url: https://codereview.chromium.org/2328263004

Remove unneeded GrGLSLTransformedCoordsArray type

Rename GrGLSLFragmentBuilder::ensureFSCoords2D to ensureCoords2D and make it take an arbitrary GrShaderVar.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2324663004

Review-Url: https://codereview.chromium.org/2324663004

Revert of Turned on SkSL->GLSL compiler (patchset #12 id:220001 of https://codereview.chromium.org/2288033003/ )

Reason for revert:
Compilation errors on some bots.

Original issue's description:
> Turned on SkSL->GLSL compiler
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003
>
> Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3

TBR=benjaminwagner@google.com,bsalomon@google.com,egdaniel@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/2337553002

Remove unneeded bounds outset in GrDrawContext::drawVertices.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2329403004

Review-Url: https://codereview.chromium.org/2329403004

glTexStorage requires sized formats, ensure that we're following that rule

In some situations (at least ANGLE ES3), we were passing GL_BGRA, rather
than GL_BGRA8.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2336613002

Review-Url: https://codereview.chromium.org/2336613002

Remove ability to skip coord transform comparison in pipeline comparison.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2316413003

Review-Url: https://codereview.chromium.org/2316413003

SkPDF: refactor & code cleanup ahead of https://crrev.com/2322403002

SkPDFDevice::GraphicStateEntry: remove unnecessary fFont and
fTextSize.

SkPDFDevice::updateFont(): replace with update_font() and inlined
code.  De-duplicate this block of code.

SkPDFResourceDict::GetResourceTypePrefix function made public: removes
need for temporary SkString returned by
SkPDFResourceDict::getResourceName()

GlyphPositioner: delay writing intial matrix until first glyph.
Assert that widechars is a constant.

SkPDFFont::FontType(): make public so that PDFDevice can know about
multibyte status.

SkPDFFont::countStretch() removed, and the stretch loop flattened.

*no changes in PDF output*

BUG=skia:5434
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2327953002

Review-Url: https://codereview.chromium.org/2327953002

Two changes:

1. Remove special premul handling from gamut xform code

Alpha is a constant, so the gamut transformation results remain unchanged
(it distributes across the linear matrix multiply).

2. Use SkMatrix44 rather than array of floats

Preserves semantic intention, and makes upcoming code (where we transform
colors on the CPU by that matrix) simpler.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2329553002

Review-Url: https://codereview.chromium.org/2329553002

Turned on SkSL->GLSL compiler
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003

Review-Url: https://codereview.chromium.org/2288033003

Update SampleAndroidShadows to use algorithm closer to Android OpenGL

Includes:
* Update light position to be at a similar distance to Android OS
* Scale spot shadows correctly
* Compute stroke shapes and radii correctly
* Allow for larger blur radius for shadows

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2319003003
TBR=reed@google.com
NOTRY=true
NOTREECHECKS=true

Review-Url: https://codereview.chromium.org/2319003003

helper to SkRect to create from SkISize

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2332733002

TBR=
NOTREECHECKS=True
NOTRY=True

Review-Url: https://codereview.chromium.org/2332733002

check for identity in translate

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2331943002

TBR=
NOTREECHECKS=True
NOTRY=True

Review-Url: https://codereview.chromium.org/2331943002

fix code behind SK_SUPPORT_LEGACY_MINOR_EFFECT_PTR

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2337433002

TBR=
NOTREECHECKS=True
NOTRY=True

Review-Url: https://codereview.chromium.org/2337433002

use expected name for setDrawLooper/getDrawLooper

Needed for future pipe cl.
Next cl should be to guard the looper entirely (since its deprecated)

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2326173002

TBR=

Review-Url: https://codereview.chromium.org/2326173002

Update SKP version

Automatic commit by the RecreateSKPs bot.

TBR=
NO_MERGE_BUILDS
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2334483002

Review-Url: https://codereview.chromium.org/2334483002

SkPDF: SkShaper_primitive uses new textblob API

This will enable me to test the new API in unit tests without
depending on HarfBuzz (after https://crrev.com/2322403002 lands).

TBR=
BUG=skia:5434

Review-Url: https://codereview.chromium.org/2332473003

Adds missing and removes duplicate tracing macros.

This fixes the missing tracing nested instant call without any arguments and
removes the duplicate of the one with 2 arguments.

This is a follow up on https://crrev.com/2326483004

BUG=645308
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2321283002
NOPRESUBMIT=true
CQ_EXCLUDE_TRYBOTS=client.skia.fyi:skia_presubmit-Trybot

Review-Url: https://codereview.chromium.org/2321283002

made point lights linear attenuation; also fixed point light depth bug

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2323383002

Review-Url: https://codereview.chromium.org/2323383002

SkPDF/Tests: imporve test coverage.

Also: make sure that all SkPDF unit tests are named SkPDF_* to
make testing changes to SkPDF easier.  Other cleanup.

Add test: SkPDF_pdfa_document to verify that flag in public API
works.

SkPDF_JpegIdentification test: test slightly malformed JPEGs to
verify that all code paths work.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2322133003

Review-Url: https://codereview.chromium.org/2322133003

Fix color gamut xform matrices in GPU code

Now that we're not storing them transposed, we can plumb them to the
shader column-major all the way, and then multiply colors on the right, as
it should be.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2326093002

Review-Url: https://codereview.chromium.org/2326093002

GN: port a few missing Android bits from GYP.

The SK_GAMMA_ defines change text rendering on Android.
This is how they're set in GYP for Android... everyone else is default (sRGB).

Linking AndroidSkDebugToStdOut.o causes SkDebugf to go to both logcat and
stdout on Android.  Without it, it only goes to logcat.  The file has no effect
on non-Android platforms.

BUG=skia:

Review-Url: https://codereview.chromium.org/2323353002

Fix storage of gamut transform matrices in SkColorSpace

We were effectively storing the transpose, which made all of our
operations on individual colors, and our concatenation of matrices
awkward and backwards.

I'm planning to push this further into Ganesh, where I had incorrectly
adjusted to the previous layout, treating colors as row vectors in the
shaders.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2324843003

Review-Url: https://codereview.chromium.org/2324843003

Skip adding coincident edges found
during curve intersection if their
ends are nearly the same.

Loosen conic/line intersection point
check.

Detect when coincident points are
unordered. This means that points
a/b/c on one curve may appear in
b/c/a order on the opposite curve.

Restructure addMissing to return
success and return if a coincidence
was added as a parameter.

With this, tiger part a works.
Tiger part b exposes bugs around
tight quads that are nearly coincident
with themselves, and are coincident
with something else.

The greedy coicident matcher
may cause the point order to be
out of sync.

Still working out what to do in
this case.

TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2321773002

Review-Url: https://codereview.chromium.org/2321773002

Added in Radial Shadows

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2311223004

Review-Url: https://codereview.chromium.org/2311223004

Delete some LUTs from SkColorSpaceXform

No need to keep these around.  In the rare case that these are
needed, we can just go ahead and build them.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2326043002

Review-Url: https://codereview.chromium.org/2326043002

fix fuzzer

Earlier fix aborted coincident loop on wrong condition.
This aborts only when it reaches the end, and then
propagates the error out correctly.

TBR=reed@google.com
BUG=644680, 644684
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2327053002

Review-Url: https://codereview.chromium.org/2327053002

Fix issues in assets scripts

- Delete unconditionally uses 'gsutil'
- args.gsutil not plumbed properly

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2326663002

Review-Url: https://codereview.chromium.org/2326663002

SkPDF: implement drawTextBlob()

Pass new fields from https://skia.googlesource.com/skia/+/4f0a23a into
SkPDFDevice::internalDrawText() (a later CL will use
them for /ActualText.

No change in PDF output.

BUG=skia:5434

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2300393002

Review-Url: https://codereview.chromium.org/2300393002

SkPDF: hand SfntlyWrapper::SubsetFont() ttcIndex, not fontName.

Also, minor code refactoring.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2258233002

Committed: https://skia.googlesource.com/skia/+/fce190647285423bf36c44bca09db78a6af30f9f
Review-Url: https://codereview.chromium.org/2258233002

Fix strange failures on Android N

We were seeing problems where dm ended with exit code 0, but the step still failed.
It seems they changed what N returns when a thing completes.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2319843003
TBR=bugsbunny

Review-Url: https://codereview.chromium.org/2319843003

Benchmarks for testing changes to GrResourceCache replacement policy when over budget.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2317263002

Review-Url: https://codereview.chromium.org/2317263002

Checking for valid colorType, alphaType, colorSpace in SkCodec

* Refactor to share code between SkPngCodec and SkWebpCodec
* Didn't end up sharing with SkJpegCodec but did refactor
  that code a bit
* Disallow conversions to F16 with non-linear color spaces
* Fail to decode if we fail to create a SkColorSpaceXform
  (should be an assert soon).  We used to fallback on a
  legacy decode if we failed to create the transform.
* A bunch of name changes

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2319293003

Committed: https://skia.googlesource.com/skia/+/7a9900d6d34e437bb24beb5524a1f6488ae138c9
Review-Url: https://codereview.chromium.org/2319293003

Move prepareForExternalIO implementation from GrContext to GrDrawingManager
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2324473002

Review-Url: https://codereview.chromium.org/2324473002

Apple devices do not support CRC32 instructions.  Don't believe Clang's lies.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2322033002
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review-Url: https://codereview.chromium.org/2322033002

Revert of Checking for valid colorType, alphaType, colorSpace in SkCodec (patchset #2 id:100001 of https://codereview.chromium.org/2319293003/ )

Reason for revert:
Broken perf bots

Original issue's description:
> Checking for valid colorType, alphaType, colorSpace in SkCodec
>
> * Refactor to share code between SkPngCodec and SkWebpCodec
> * Didn't end up sharing with SkJpegCodec but did refactor
>   that code a bit
> * Disallow conversions to F16 with non-linear color spaces
> * Fail to decode if we fail to create a SkColorSpaceXform
>   (should be an assert soon).  We used to fallback on a
>   legacy decode if we failed to create the transform.
> * A bunch of name changes
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2319293003
>
> Committed: https://skia.googlesource.com/skia/+/7a9900d6d34e437bb24beb5524a1f6488ae138c9

TBR=scroggo@google.com,brianosman@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/2328663002

Checking for valid colorType, alphaType, colorSpace in SkCodec

* Refactor to share code between SkPngCodec and SkWebpCodec
* Didn't end up sharing with SkJpegCodec but did refactor
  that code a bit
* Disallow conversions to F16 with non-linear color spaces
* Fail to decode if we fail to create a SkColorSpaceXform
  (should be an assert soon).  We used to fallback on a
  legacy decode if we failed to create the transform.
* A bunch of name changes

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2319293003

Review-Url: https://codereview.chromium.org/2319293003

Cache the inverse matrix on SkColorSpace. Rename xyz() to toXYZ().

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2323003002

Review-Url: https://codereview.chromium.org/2323003002

GN: add a helper script for running Android builds

Think of this as the spiritual equivalent of platform_tools/android/bin/android_run_skia, but for GN and easier to type.

Cutting down the debug symbols makes Android builds about 1/4 the size.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2323833002
DOCS_PREVIEW= https://skia.org/user/quick/gn?cl=2323833002

Review-Url: https://codereview.chromium.org/2323833002

Push usage of GrColor4f into OverrideInput

Just a change to preserve precision.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2324553002

Review-Url: https://codereview.chromium.org/2324553002

most is fine for *SAN bots now

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2326553002

Review-Url: https://codereview.chromium.org/2326553002

removed const from ASTVarDeclaration::fName to avoid a Chromium build failure
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2325673002

Review-Url: https://codereview.chromium.org/2325673002

Clean up target_os=llvm now that it does nothing.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2317373004

Review-Url: https://codereview.chromium.org/2317373004

Add GM/slide to simulate Android-style reveal clip

Hopefully, this will let us play w/ geometric and shader-based solutions.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2316593003

Committed: https://skia.googlesource.com/skia/+/ffac5c4aae18fc706e4077763c190a89c8507fb0
Review-Url: https://codereview.chromium.org/2316593003

Tests: DiscardableMemory test no longer relies on global state

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2323013002

Review-Url: https://codereview.chromium.org/2323013002

SkLiteDL: drawAsLayer()

This more or less the API you were thinking?

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2315283002

Review-Url: https://codereview.chromium.org/2315283002

GN: add sanitize arg

Attempt to take over all *SAN builds.

MSAN has a lot of coordination required between gn/BUILD.gn and gn_flavor.py.
I'd like to follow up to move more of this into gn/BUILD.gn, to make it easier
to use locally.

The compile steps should be much faster now.  We no longer build CMake
and Clang for every run, instead using the clang_linux CIPD package.  This
removes the need for all the third_party/externals/llvm/... dependencies.

Similarly, since we're using the clang_linux package, we no longer depend
on Chrome's Clang, and thus no longer need to sync chromium on these bots.

Instead of packaging up MSAN libraries and llvm-symbolizer in the compile
output, I have the test / perf bots also depend on the clang_linux package.
These do not vary from build to build.

No more need for the xsan.blacklist -include hack: Clang, GN, and Ninja
all track changes to xsan.blacklist without our help.

This has the incidental effect of upgrading the compiler used by *SAN
bots from Clang 3.8 to Clang 3.9.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2289343002

Review-Url: https://codereview.chromium.org/2289343002

Flesh out SkOpts namespaces.

This makes it easier to see the baseline CPU feature settings for a build when looking at a stack trace or profile.  This fills out all the features we currently
might care about.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2318313003

Review-Url: https://codereview.chromium.org/2318313003

Asset management: Download from GS store instead of CIPD

This avoids strange CIPD symlink/permissions issues

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2321623003

Review-Url: https://codereview.chromium.org/2321623003

GN: do not set default GN args

It's just visual noise... this should make future gn_flavor.py
changes show fewer incidental diffs.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2324573002

Review-Url: https://codereview.chromium.org/2324573002

Further increase run_benchmark timeout

Looks like the chromium page sets still hit the 10 min timeout.

BUG=skia:5700

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2342
# This code is not exercised by trybots.
NOTRY=true

Change-Id: I57aba4c1541a840d7e4e97112a6485abc20aa3c7
Reviewed-on: https://skia-review.googlesource.com/2342
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>

Add color xform support to SkWebpCodec

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2294993002

Committed: https://skia.googlesource.com/skia/+/828ed1777da74692d0c8a5834017929f5aedcc6b
Review-Url: https://codereview.chromium.org/2294993002

Delete deprecated GammaNamed

TBR=reed@google.com
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2320893002

Review-Url: https://codereview.chromium.org/2320893002

GN: default to -O1

In a GN world we are no longer tied to the Debug/Release dichotomy of
GYP, and extra_cflags makes it trivially easy to override any flags we
set by default.  These two together mean we don't have to make the default
Debug build pointlessly unoptimized.

-O1 takes negligibly longer to compile than -O0, doesn't make debugging
difficult, and the code is considerably better.  DM runs ~4x faster this way.

Backstory: I just re-noticed we do this for our Debug MSAN builds, which has
had a serious wall-time slimming effect.  I was about to restore it to my GN
MSAN builds when I realized, hey, maybe we should do this for all builds.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2322753002

Review-Url: https://codereview.chromium.org/2322753002

Rename SkColorSpace::GammaNamed to SkColorSpace::RenderTargetGamma

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2319293002

Review-Url: https://codereview.chromium.org/2319293002

Revert of Add color xform support to SkWebpCodec (patchset #2 id:80001 of https://codereview.chromium.org/2294993002/ )

Reason for revert:
Windows bots hate it when I upload new images.

Original issue's description:
> Add color xform support to SkWebpCodec
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2294993002
>
> Committed: https://skia.googlesource.com/skia/+/828ed1777da74692d0c8a5834017929f5aedcc6b

TBR=scroggo@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/2318253004