remove legacy moz flags

NOTRY=True
R=george@mozilla.com, scroggo@google.com

Author: reed@google.com

Review URL: https://codereview.chromium.org/342803002

Modify gpu debug markers to store gpu device markers when a flush occurs.

BUG=skia:
R=bsalomon@google.com

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/341583002

Remove unused class from the Android framework build

R=reed@google.com, scroggo@google.com

Author: djsollen@google.com

Review URL: https://codereview.chromium.org/345613002

No CrashHandler for NACL bots.

CQ_EXTRA_TRYBOTS=tryserver.skia:Build-Ubuntu13.10-GCC4.8-NaCl-Debug-Trybot

BUG=skia:
R=mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/344603005

Add basic stacktrace handler using libunwind.

This means we will all have to apt-get install libunwind8-dev on Linux.  Mac comes with everything we need already.

BUG=skia:
R=reed@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/343583005

In Android framework, make tools depend on jsoncpp

Always build the tools with JSON, but either build our own
or use the system's.

Rename skia_build_json_writer to skia_use_system_jsoncpp,
since we now always build with JSON.

Remove SK_BUILD_JSON_WRITER, which was only there so
we could build without JSON it in the framework.

BUG=skia:2448
R=djsollen@google.com, reed@google.com

Author: scroggo@google.com

Review URL: https://codereview.chromium.org/303913002

Add SkASSERTF.

Example failure:
fRefCnt was 3
../../../usr/local/google/home/mtklein/skia/include/core/SkRefCnt.h:40: failed assertion "(fRefCnt == 1) || (SkDebugf("fRefCnt was %d""\n", fRefCnt), false)"
Command terminated by signal 11

Not pretty, but everything's there.  Perhaps we'll think of ways to make it nicer later.

BUG=skia:
R=bsalomon@google.com, reed@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/337243004

Remove dashing from gpu veto

With new veto our new veto test results look like the following:
TP: true positive (picked to use gpu and gpu was faster)
I: inderminate, the raster time is withing 5% of gpu time

        TP  FP  TN  FN  I

old     21  9   15  12  3
new     29  12  11  6   3

There are three skps that tend to move from TN -> FP, however
the absolute difference in their run times are not huge between
them. The largest being desk_booking which is about 7.1 raster
and 8.8 gpu. The other two skps are desk_yahooanswers and
desk_linkedin

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

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/334053005

Add additional dashing rebaselines
NOTRY=True
NOTREECHECK=True
R=bsalomon@google.com
TBR=bsalomon@google.com

BUG=skia:

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/347543003

Fix last pathops skp bug

This fixes the last bug discovered by iterating through the 800K
skp corpus representing the top 1M websites. For every clip on the
stack, the paths are replaced with the pathop intersection. The
resulting draw is compared with the original draw for pixel errors.

At least two prominent bugs remain. In one, the winding value is
confused by a cubic with an inflection. In the other, a quad/cubic
pair, nearly coincident, fails to find an intersection.

These minor changes include ignoring very tiny self-intersections
of cubics, and processing degenerate edges that don't connect to
anything else.

R=reed@android.com
TBR=reed

Author: caryclark@google.com

Review URL: https://codereview.chromium.org/340103002

Remove SkBenchmark::postDraw.

Only one bench used this feature, and it's for things that'd be done by the
destructor anyway.

BUG=skia:
R=bsalomon@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/329853007

Use DNS and not hard coded IP address of the monitoring server.

BUG=skia:
R=mtklein@google.com

Author: jcgregorio@google.com

Review URL: https://codereview.chromium.org/338873003

Rebaslines from recent dashing changes
R=bsalomon@google.com
TBR=bsalomon@google.com
BUG=skia:

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/339203003

Removes Record configs from buildbot bench_pictures runs.

BUG=skia:
NOTRY=true
R=borenet@google.com, mtklein@google.com

Author: bensong@google.com

Review URL: https://codereview.chromium.org/338413002

unignore canvas-layer-state

TBR=
NOTRY=True

Author: reed@google.com

Review URL: https://codereview.chromium.org/336273005

Update SKP version to 26

Automatic commit by the RecreateSKPs bot.

TBR=

Author: borenet@google.com

Review URL: https://codereview.chromium.org/342653002

Clean up DirectWrite typeface cache matching.

The matching code was difficult to follow due to naming issues,
and performed some duplicate work which is not wanted.
This change will either fix the associated bug or make it possible
to track the cause.

CQ_EXTRA_TRYBOTS=tryserver.skia:Test-Win7-ShuttleA-HD2000-x86-Release-DirectWrite-Trybot

BUG=384529
R=reed@google.com

Author: bungeman@google.com

Review URL: https://codereview.chromium.org/338923003

Revert of Add SSE4 optimization of S32A_Opaque_Blitrow (https://codereview.chromium.org/289473009/)

NOTREECHECKS=true
NOTRY=true

Reason for revert:
Valgrind bot's seeing this code use uninitialized memory, and it's somehow blocking our roll into Chrome too:

> ld: warning: could not create compact unwind for
S32A_Opaque_BlitRow32_SSE4_asm:
> stack subq instruction is too different from dwarf stack size
> [10339/10982 | 3247.792] PACKAGE FRAMEWORK "Chromium Framework.framework",
> POSTBUILDS
> FAILED: ./gyp-mac-tool package-framework "Chromium Framework.framework" A &&
> (export
> BUILT_PRODUCTS_DIR=/Volumes/data/b/build/slave/mac_gpu/build/src/out/Release;
> export CONFIGURATION=Release; export CONTENTS_FOLDER_PATH="Chromium
> Framework.framework/Versions/A"; export
> DYLIB_INSTALL_NAME_BASE=@executable_path/../Versions/37.0.2056.0; export
> EXECUTABLE_NAME="Chromium Framework"; export EXECUTABLE_PATH="Chromium
> Framework.framework/Versions/A/Chromium Framework"; export
> FULL_PRODUCT_NAME="Chromium Framework.framework"; export
> INFOPLIST_PATH="Chromium Framework.framework/Versions/A/Resources/Info.plist";
> export
LD_DYLIB_INSTALL_NAME="@executable_path/../Versions/37.0.2056.0/Chromium
> Framework.framework/Chromium Framework"; export MACH_O_TYPE=mh_dylib; export
> PRODUCT_NAME="Chromium Framework"; export
> PRODUCT_TYPE=com.apple.product-type.framework; export
>
SDKROOT=/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.6.sdk;
> export
>
SRCROOT=/Volumes/data/b/build/slave/mac_gpu/build/src/out/Release/../../chrome;
> export SOURCE_ROOT="${SRCROOT}"; export
> TARGET_BUILD_DIR=/Volumes/data/b/build/slave/mac_gpu/build/src/out/Release;
> export TEMP_DIR="${TMPDIR}"; export
UNLOCALIZED_RESOURCES_FOLDER_PATH="Chromium
> Framework.framework/Versions/A/Resources"; export WRAPPER_NAME="Chromium
> Framework.framework"; (cd ../../chrome && ../build/mac/tweak_info_plist.py
> "--breakpad=1" "--breakpad_uploads=0" "--keystone=0" "--scm=1"
> "--branding=Chromium" && ln -fns Versions/Current/Libraries
> "${BUILT_PRODUCTS_DIR}/${WRAPPER_NAME}/Libraries" &&
> tools/build/mac/verify_order _ChromeMain
> "${BUILT_PRODUCTS_DIR}/${EXECUTABLE_PATH}"); G=$?; ((exit $G) || rm -rf
> 'Chromium Framework.framework') && exit $G) && touch "Chromium
> Framework.framework"
> tools/build/mac/verify_order: unordered symbols in
> /Volumes/data/b/build/slave/mac_gpu/build/src/out/Release/Chromium
> Framework.framework/Versions/A/Chromium Framework:
> S32A_Opaque_BlitRow32_SSE4_asm
> _S32A_Opaque_BlitRow32_SSE4_asm
> ninja: build stopped: subcommand failed.

Original issue's description:
> Add SSE4 optimization of S32A_Opaque_Blitrow
>
> Adds optimization of Skia S32A_Opaque_Blitrow blitter using SSE4.2 SIMD
> instruction set. Special case for when alpha is zero or opaque.
>
> Performance increase of 10%-400% compared to the existing SSE2
> optimization (measured on Silvermont architecture).
> Noticeable in ~25 different skia bench subtests, especially in
> bitmap_8888_*, repeatTile_*, and morph_*.
>
> bitmap_8888_A - 100% faster
> bitmap_8888_A_source_transparent - 250% faster
> bitmap_8888_A_source_opaque - 25% faster
> bitmap_8888_A_scale_bicubic - 75% faster
>
> Signed-off-by: Henrik Smiding <henrik.smiding@intel.com>
>
> Committed: https://skia.googlesource.com/skia/+/e2527b147679b0c43019fae7d59cc3777d2d097e
>
> Committed: https://skia.googlesource.com/skia/+/b5c281e1e06af3be804309877de1dac6145686b9

R=reed@google.com, tomhudson@google.com, djsollen@google.com, joakim.landberg@intel.com, henrik.smiding@intel.com, mtklein@chromium.org

Author: mtklein@google.com

Review URL: https://codereview.chromium.org/336413007

Rebaseline ANGLE bots.

Try to roll angle again.

BUG=skia:2272

NOTRY=True
NOTREECHECKS=True
R=reed@google.com

Review URL: https://codereview.chromium.org/306483014

More SaveFlags removal twiddling.

This is a follow up to https://codereview.chromium.org/338913002/. More
tweaks to allow migrating clients to the new API.

1) switch the shim call direction willSave(SaveFlags) -> willSave()
   (internal users are still using the former, so external overriders of
    the latter will not be notified otherwise - doh)
2) ensure willSave() stays visible in SkProxyCanvas (Chromium's
   TimingCanvas attempts to call it explicitly).

BUG=skia:2297
R=reed@google.com

Author: fmalita@chromium.org

Review URL: https://codereview.chromium.org/334393008

Update the name of the friended android class and add the appropriate guards.

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

Author: djsollen@google.com

Review URL: https://codereview.chromium.org/344473002

Fifth attempt to land faster rect blur.  Remove unnecessary conditionals from the shader.

approved over in https://codereview.chromium.org/331863006/; reuploading
because it had the wrong base url

BUG=skia:2095
R=bsalomon@google.com
TBR=bsalomon

Author: humper@google.com

Review URL: https://codereview.chromium.org/341543005

Update Windows baselines due to VS2013 update.

Floating point operations now differ in the low bits from before.

rebaseline after changing labels in bitmapfilters

TBR=
NOTRY=True
NOTREECHECKS=True

Author: reed@google.com

Review URL: https://codereview.chromium.org/340583004

Add EXPERIMENTAL_beginRecording() for SkRecord-based recording.

The interesting stuff is in SkPictureRecorder.{h,cpp}.  The rest is mostly moving SkRecord from its own directories into core to avoid circular dependencies in GYP.

After plumbing SkRecord all the way through in Picture, I'll delete its old entry point include/record/SkRecording.h.  For now it and record.gypi need to stay where they are to keep Chrome building.

BUG=skia:
R=reed@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/331573004

Silence warnings about using offsetof on near-POD structs.

This is coming up with SkMatrix, which is POD according to GCC >= 4.6
but not before (specifically, not at GCC 4.4).

BUG=skia:
R=reed@google.com, mtklein@google.com, bungeman@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/339193002

Add SSE4 optimization of S32A_Opaque_Blitrow

Adds optimization of Skia S32A_Opaque_Blitrow blitter using SSE4.2 SIMD
instruction set. Special case for when alpha is zero or opaque.

Performance increase of 10%-400% compared to the existing SSE2
optimization (measured on Silvermont architecture).
Noticeable in ~25 different skia bench subtests, especially in
bitmap_8888_*, repeatTile_*, and morph_*.

bitmap_8888_A - 100% faster
bitmap_8888_A_source_transparent - 250% faster
bitmap_8888_A_source_opaque - 25% faster
bitmap_8888_A_scale_bicubic - 75% faster

Signed-off-by: Henrik Smiding <henrik.smiding@intel.com>
Committed: https://skia.googlesource.com/skia/+/e2527b147679b0c43019fae7d59cc3777d2d097e

R=reed@google.com, mtklein@google.com, tomhudson@google.com, djsollen@google.com, joakim.landberg@intel.com

Author: henrik.smiding@intel.com

Review URL: https://codereview.chromium.org/289473009

* Removed bitmapshaders from expectations/gm/ignored-tests.txt.
* Rebaseline bitmapshaders, gpu didn't produce the same images when using only alpha bitmaps
because it didn't consider the skpaint's color.

BUG=skia:2293
R=jvanverth@google.com

Author: dandov@google.com

Review URL: https://codereview.chromium.org/339923004

DirectWrite to use PostScript name for PDF.

The PDF backend expects info->fFontName to be the PostScript name.
Prior to this change the '<family> <sytle>' name was used instead.

R=vandebo@chromium.org, halcanary@google.com, reed@google.com, thestig@chromium.org

Author: bungeman@google.com

Review URL: https://codereview.chromium.org/324403010

rebaseline after GM change to bitmapfilters

TBR=
NOTRY=True
NOTREECHECKS=True

Author: reed@google.com

Review URL: https://codereview.chromium.org/341593003

ignore bitmapfilters failures until we can rebaseline

TBR=
NOTRY=True
NOTREECHECKS=True

Author: reed@google.com

Review URL: https://codereview.chromium.org/339153004

move some headers out of public

patch from issue 338263003

BUG=skia:
R=mtklein@google.com

Author: reed@google.com

Review URL: https://codereview.chromium.org/339183002

hide SkBitmap::Config entirely (behind a flag)

patch from issue 339463002

TBR=

I think the NoGPU failure is unrelated, so ignoring
NOTRY=True

Author: reed@google.com

Review URL: https://codereview.chromium.org/340533002

Move SkStringUtils.h to src/core.

Doesn't need to be in include/.

BUG=skia:
R=reed@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/335413004

BenchSysTimer_mach drops all sub-ms precision.  Fix that!

BUG=skia:
R=bungeman@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/336903005

fix chrome
had #if SK_DEBUG instead of #ifdef SK_DEBUG

R=reed@android.com
TBR=reed
NOTRY=true
NOTREECHECKS=true

Author: caryclark@google.com

Review URL: https://codereview.chromium.org/338333002

Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites.

This fixes all but one of those failures.

Major changes include:
- Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles.
- Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed.
- Line segments with ends nearly the same are treated as coincident first.
- Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident.

TBR=reed

Author: caryclark@google.com

Review URL: https://codereview.chromium.org/272153002

Add nanomsg to third_party, with some demos.

This isn't something I want to make part of Skia, but just a substrate to build cross-process demos on top of.  If I client were to use Skia cross-process, they'd drop their own IPC system in here.

If you're not familiar, nanomsg (nanomsg.org) is the next-gen zeromq (zeromq.org), from the same author, righting all his design wrongs from zeromq.

It's a lot like the lower half of mojo, dealing with making the connections and getting messages reliably from A to B.  Think, better sockets, and it spans nicely across in-process (with zero-copy), inter-process, and TCP.

BUG=skia:
R=bsalomon@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/294873004

Split out DEPS change from 294873004 to land first.

BUG=skia:
R=mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/331913004

remove SK_SUPPORT_LEGACY_BLURMASKFILTER_STYLE -- not needed

BUG=skia:
R=scroggo@google.com

Author: reed@google.com

Review URL: https://codereview.chromium.org/332163005

Refine bench_record and bench_playback:

  - use high-precision wall timer only
  - warm caches once before measuring
  - measure independent samples, calculating statistics
  - add --verbose to control how much data we output

Also removed some unloved features from bench_record.

BUG=skia:
R=jcgregorio@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/338203002

Fix Windows bots

BUG=skia:
R=mtklein@google.com, rmistry@google.com

Author: borenet@google.com

Review URL: https://codereview.chromium.org/339733002

Revert of Temporarily limit x86 SIMD to SSE2 only, to see effect on all benches and bots. (https://codereview.chromium.org/331193004/)

Reason for revert:
Experiment is over: disabling SSSE3 is a 25-50% perf regression for bitmap scaling on every machine we've got.

Original issue's description:
> Temporarily limit x86 SIMD to SSE2 only, to see effect on all benches and bots.
>
> BUG=372232
>
> Committed: https://skia.googlesource.com/skia/+/f1e5a04832e4d350f9ebf5d556c6d3897345f883

R=reed@google.com, mtklein@chromium.org
TBR=mtklein@chromium.org, reed@google.com
NOTREECHECKS=true
NOTRY=true
BUG=372232

Author: mtklein@google.com

Review URL: https://codereview.chromium.org/332213005

Prepare for internal SaveFlags removal.

Chromium has a couple of SkCanvas subclasses which override willSave().
This adds a transitional shim to facilitate converting these to the
parameter-less API.

BUG=skia:2297
R=reed@google.com, scroggo@google.com

Author: fmalita@chromium.org

Review URL: https://codereview.chromium.org/338913002

Cleaned a repeated if(condition) with the same condition

R=junov@chromium.org

Author: piotaixr@chromium.org

Review URL: https://codereview.chromium.org/338933002

Perlin noise cleanup.

Remove Simplex noise implementation and now-redundant base classes.

BUG=skia:
R=sugoi@chromium.org, sugoi@google.com, reed@google.com

Author: senorblanco@chromium.org

Review URL: https://codereview.chromium.org/336043002

Temporarily limit x86 SIMD to SSE2 only, to see effect on all benches and bots.

BUG=372232
R=reed@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/331193004

Armv8/64bit doesn't (apparently) need rebaselining

BUG=skia:
R=reed@google.com
NOTRY=True

Author: kevin.petit@arm.com

Review URL: https://codereview.chromium.org/334743009

Add sugoi@ to CQ_COMMITTERS.

R=junov@chromium.org

Author: senorblanco@chromium.org

Review URL: https://codereview.chromium.org/337033005

fix legacy code path for SK_SUPPORT_LEGACY_IMAGEDECODER_CONFIG

BUG=skia:
R=scroggo@google.com

Author: reed@google.com

Review URL: https://codereview.chromium.org/332223002

Enable compiling with automatically-downloaded VS toolchain on Windows bots

Requires that the bots have CHROME_HEADLESS=1 and CHROME_PATH set.

BUG=skia:2216
R=epoger@google.com

Review URL: https://codereview.chromium.org/332833002

Update SKP version to 24

Automatic commit by the RecreateSKPs bot.

TBR=

Author: borenet@google.com

Review URL: https://codereview.chromium.org/332983003

Minor cleanups in picture_utils header file.

1) Remove unused include.
2) Remove unused forward declarations.

BUG=None
TEST=make tests
R=epoger@google.com
TBR=epoger@google.com

Author: tfarina@chromium.org

Review URL: https://codereview.chromium.org/339493004

Update SKP version to 23

Automatic commit by the RecreateSKPs bot.

TBR=

Author: borenet@google.com

Review URL: https://codereview.chromium.org/332843006

stop using SkBitmap::Config

R=scroggo@google.com

Author: reed@chromium.org

Review URL: https://codereview.chromium.org/338493005

Update SKP version to 22

Automatic commit by the RecreateSKPs bot.

TBR=

Author: borenet@google.com

Review URL: https://codereview.chromium.org/332013002

Delete unused path_utils.* component.

filter tool includes it but isn't using, and nobody else seems to
be using it either. So just remove it now.

BUG=None
TEST=make filter
R=mtklein@google.com

Author: tfarina@chromium.org

Review URL: https://codereview.chromium.org/324273002

hide virtuals on device for width/height/isopaque

R=bsalomon@google.com
NOTREECHECKS=True

Author: reed@chromium.org

Review URL: https://codereview.chromium.org/334993002

Testing trybot triggering

CQ_EXTRA_TRYBOTS=tryserver.skia:Build-Ubuntu13.10-GCC4.8-x86_64-Debug-CrOS_Link-Trybot
BUG=skia:2659
NOTREECHECKS=true
TBR=
(SkipBuildbotRuns)

Author: rmistry@google.com

Review URL: https://codereview.chromium.org/332933005

Extract "text align proc" functions as reusable classes

Extract "text align proc" as reusable classes. These classes need to be used
when writing GrTextContext subclasses.

Moves "text align proc" code that is duplicated in SkDraw and
SkBitmapTextContext to SkDrawProcs.h and SkTextMapState.h. This functionality is
also used in the new GrStencilAndCoverTextContext.

Creates new functor classes SkTextAlignProc and SkTextAlignProcScalar which
represent the previous "text align procs".

Moves TextMapState from SkDraw to SkTextMapStateProc and make it similar functor.

The transform should be comparable in speed, as the compiler can and does avoid
the call and eliminate some of the branches.

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

Author: kkinnunen@nvidia.com

Review URL: https://codereview.chromium.org/335573002

Randomize seed for SkDiscretePathEffect::filterPath()

Fix for https://code.google.com/p/skia/issues/detail?id=2581. Randomizes
the seed in SkDiscretePathEffect::filterPath(). Prior to this we were
using the path length as a seed.

Now, if we have two different paths with identical contents and we
apply an SkDiscretePathEffect to each, we obtain two different random
paths. Previously, we would obtain two overlapping paths (identical
path contents leading to the same seed).

BUG=skia:
R=reed@google.com, scroggo@google.com

Author: rs.prinja@samsung.com

Review URL: https://codereview.chromium.org/311803002

hide Config in SkImageDecoder -- use SkColorType instead

patch from issue 334613003

TBR=scroggo

Author: reed@chromium.org

Review URL: https://codereview.chromium.org/334793002

New baselines for perlin noise tests affected by ce6a35.

New baselines for perlinnoise, imagefiltersclipped and
imagefiltersscaled GMs.

R=scroggo@google.com
TBR=scroggo
NOTRY=true

Author: senorblanco@chromium.org

Review URL: https://codereview.chromium.org/336703002

Remove SkPicture SK_SUPPORT_LEGACY_RECORDING_FLAG flag

Chromium/Blink should no longer need this flag after:

Chromium: Remove use of kUsePathBoundsForClip_RecordingFlag https://codereview.chromium.org/322123002/

Blink: Remove use of kUsePathBoundsForClip_RecordingFlag https://codereview.chromium.org/326953002/

R=mtklein@google.com, scroggo@google.com, reed@google.com, bsalomon@google.com

Author: robertphillips@google.com

Review URL: https://codereview.chromium.org/328343002

rebaselines for removal of unitmapper

TBR=
NOTRY=True

Author: reed@google.com

Review URL: https://codereview.chromium.org/330513005

Add more tests to ignore-tests.txt.

Add imagefiltersclipped and imagefiltersscaled to ignored-tests.txt
until they can be rebaselined, since they're failing on Android after
https://codereview.chromium.org/332523006/.

TBR=scroggo
BUG=skia:

Review URL: https://codereview.chromium.org/336663002

Revert of third try at landing improved blur rect; this time with more correctness (https://codereview.chromium.org/331443003/)

Reason for revert:
Failing layout test: https://storage.googleapis.com/chromium-layout-test-archives/WebKit_Linux/32762/layout-test-results/virtual/gpu/fast/canvas/canvas-draw-canvas-on-canvas-shadow-pretty-diff.html

Original issue's description:
> third try at landing improved blur rect; this time with more correctness
>
> BUG=skia:2095
> R=bsalomon@google.com
> TBR=bsalomon
>
> Committed: https://skia.googlesource.com/skia/+/72abfc2b4e7caead660f6b6a05e60d05eaf1a66f

R=bsalomon@google.com, reed@google.com, humper@google.com
TBR=bsalomon@google.com, humper@google.com, reed@google.com
NOTREECHECKS=true
NOTRY=true
BUG=skia:2095

Author: scroggo@google.com

Review URL: https://codereview.chromium.org/333763002

Add another fail-ignore for flaky dashing4 gm
NOTREECHECKS=True
NOTRY=True
TBR=bsalomon@google.com
BUG=skia:

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/329223007

Greenify bench bot Perf-Ubuntu12-ShuttleA-GTX660-x86-Release at build 968

R=bsalomon@google.com
TBR=scroggo@google.com

Bypassing trybots:
NOTRY=true

Author: bensong@google.com

Review URL: https://codereview.chromium.org/338453004

Added in framework to get more bench data

BUG=skia:
NOTRY=true
R=jcgregorio@google.com, bensong@google.com

Author: kelvinly@google.com

Review URL: https://codereview.chromium.org/331683003

Fix tiled perlin noise.

It turns out that the perlin implementation we inherited from WebKit
does not actually generate tileable noise (see Chromium bug
http://crbug.com/383495).

The main problem is that when generating coordinates for gradient
interpolation, it was attempting to wrap both x and (x + 1)
simultaneously at the tile boundary (that is, either both or neither
are wrapped). This obviously won't work, since along the tile seams,
(x + 1) should be wrapped, but x should not. The same is true in y.

This patch fixes both the CPU and GPU paths, renames some variables to
more closely match the spec, and modifies the perlin noise GM to
actually test tiling. (Note that the clipping the GM was doing was
removed, since it's superfluous: it used to be necessary for image
filters, but isn't anymore, and this isn't an image filter GM anyway.)

R=sugoi@google.com, sugoi
TBR=senorblanco

Author: senorblanco@chromium.org

Review URL: https://codereview.chromium.org/332523006

Use vertex attributes for dash effect in gpu

This will allow us to batch dashed lines together when drawing. Also, this removes the need for
a coord transform matrix in the shader, thus we save the cost of uploading a new matrix uniform
everytime we do a simple transform to the dashed line we are drawing.

BUG=skia:
R=bsalomon@google.com

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/326103002

manual bench rebase after 679426c using old algorithm

R=robertphillips@google.com
TBR=borenet@google.com, robertphillips@google.com

Bypassing trybots:
NOTRY=true

Author: bensong@google.com

Review URL: https://codereview.chromium.org/330103002

remove GetDeviceConfig/SetDeviceConfig

BUG=skia:
R=scroggo@google.com

Author: reed@google.com

Review URL: https://codereview.chromium.org/332453002

Use scratch with LATC data if possible

R=robertphillips@google.com

Author: krajcevski@google.com

Review URL: https://codereview.chromium.org/330593004

Cleanup: Delete sk_tools::make_filepath() in favor of SkOSPath::SkPathJoin().

BUG=None
TEST=make tools tests && out/Debug/tests
R=epoger@google.com

Author: tfarina@chromium.org

Review URL: https://codereview.chromium.org/327403002

removes deprecated -q flag for gsutil.

BUG=skia:
NOTRY=true
R=kelvinly@google.com
TBR=kelvinly@google.com

Author: bensong@google.com

Review URL: https://codereview.chromium.org/333643005

Really revert bench trigger changes.

BUG=skia:
NOTRY=true
R=kelvinly@google.com
TBR=kelvinly@google.com

Author: bensong@google.com

Review URL: https://codereview.chromium.org/324403009

Rebase flakey windows dash4 gm
NOTREECHECKS=True
NOTRY=True
TBR=bsalomon@google.com
BUG=skia:2667

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/335443003

Revert trigger changes

BUG=skia:
NOTRY=true
R=robertphillips@google.com, bensong@google.com, jcgregorio@google.com

Author: kelvinly@google.com

Review URL: https://codereview.chromium.org/334673002

Add expectations for dashing gm
TBR=bsalomon@google.com
BUG=skia:

Author: egdaniel@google.com

Review URL: https://codereview.chromium.org/330033002

Whitespace change to test trybot triggering

BUG=skia:
CQ_EXTRA_TRYBOTS=tryserver.skia:Build-Ubuntu13.10-GCC4.8-Arm7-Release-Android-Trybot
TBR=
(SkipBuildbotRuns)

Author: rmistry@google.com

Review URL: https://codereview.chromium.org/333663002

manual bench rebase after 679426c

In response to:
- new SKPs
- Kelvin's changes around regression testing

R=robertphillips@google.com, kelvinly@google.com
TBR=robertphillips@google.com

Bypassing trybots:
NOTRY=true

Author: scroggo@google.com

Review URL: https://codereview.chromium.org/332483004

Remove SkPicture pointer from SkPicturePlayback

This CL simplifies the relationship between SkPicture and SkPicturePlayback by moving the path heap into SkPicturePlayback and removing SkPicturePlayback's SkPicture pointer.

R=mtklein@google.com, reed@google.com

Author: robertphillips@google.com

Review URL: https://codereview.chromium.org/334493002

Update SKP version to 21

Automatic commit by the RecreateSKPs bot.

TBR=

Author: borenet@google.com

Review URL: https://codereview.chromium.org/327213004

rebaseline_server: delete no-longer-used weightedDiffMeasure
(replaced by perceptual diff)

NOTRY=True
R=rmistry@google.com
TBR=rmistry

Author: epoger@google.com

Review URL: https://codereview.chromium.org/334533003

rebaseline_server: change 2 image expectations within unittests

NOTRY=True
R=rmistry@google.com
TBR=rmistry

Author: epoger@google.com

Review URL: https://codereview.chromium.org/329933002

Suppress libpoppler and its dependent liblcms in TSAN.

This will let us eventually turn PDF back on in DM.

BUG=skia:
R=halcanary@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/330433003

Add ifdef to not use function pointer that we don't have yet on Chromium??

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

Author: krajcevski@google.com

Review URL: https://codereview.chromium.org/326423003

Allow compressed texture data to be updated.

R=robertphillips@google.com

Author: krajcevski@google.com

Review URL: https://codereview.chromium.org/326383003

rebaseline_server: merge identical results (across multiple builders/configs) into a single row

BUG=skia:2534
R=jcgregorio@google.com

Author: epoger@google.com

Review URL: https://codereview.chromium.org/326013002

Fix a number of issues with iOS build.

Set pointer to root view in window (for SampleApp)
Set up the correct function pointer for GL multisampling
Remove use of SkBitmap::setConfig()
Remove incomplete iOS implementation of SkFILE (use SkOSFile for now)

R=caryclark@google.com, reed@google.com, bsalomon@google.com, tfarina@chromium.org

Author: jvanverth@google.com

Review URL: https://codereview.chromium.org/322403007

third try at landing improved blur rect; this time with more correctness

BUG=skia:2095
R=bsalomon@google.com
TBR=bsalomon

Author: humper@google.com

Review URL: https://codereview.chromium.org/331443003

Texture compression unit tests along with a couple of bug fixes

R=robertphillips@google.com

Author: krajcevski@google.com

Review URL: https://codereview.chromium.org/330453005

SkNEW as default Create() for SkLazyPtr macros.

BUG=skia:
R=halcanary@google.com, mtklein@google.com

Author: mtklein@chromium.org

Review URL: https://codereview.chromium.org/326413003

Remove picture pre-allocation from SkPictureRecorder

This CL improves the separation of the SkPicture and SkPictureRecord classes. It delays creation of the SkPicture (in SkPictureRecorder) until recording is actually completed. To accomplish this the SkRecord-derived classes now get SkPathHeap and SkPictureContentInfo members that are absorbed by the SkPicture when it is constructed.

As an ancillary change, this CL also moves the SkPictureContentInfo object from SkPicture to SkPicturePlayback. This is intended to centralize all the data in the SkPicturePlayback object.

R=mtklein@google.com, reed@google.com

Author: robertphillips@google.com

Review URL: https://codereview.chromium.org/324293004

Hide chooseFromOneChoice along with Chooser.

Fixes chrome canary.

R=reed@android.com
TBR=reed
NOTRY=true

Author: scroggo@google.com

Review URL: https://codereview.chromium.org/329113003

Add support for glCompressedTexSubImage2D

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

Author: krajcevski@google.com

Review URL: https://codereview.chromium.org/329213002

hide SkImageDecoder::Chooser

BUG=skia:
R=scroggo@google.com

Author: reed@google.com

Review URL: https://codereview.chromium.org/331433003