skia2/tools/dm_flags.py
reed 320a40d773 Always return ImageShader, even from SkShader::MakeBitmapShader
Lessons learned

1. ImageShader (correctly) always compresses (typically via PNG) during serialization. This has the surprise results of
- if the image was marked opaque, but has some non-opaque pixels (i.e. bug in blitter or caller), then compressing may "fix" those pixels, making the deserialized version draw differently. bug filed.
- 565 compressess/decompresses to 8888 (at least on Mac), which draws differently (esp. under some filters). bug filed.

2. BitmapShader did not enforce a copy for mutable bitmaps, but ImageShader does (since it creates an Image). Thus the former would see subsequent changes to the pixels after shader creation, while the latter does not, hence the change to the BlitRow test to avoid this modify-after-create pattern. I sure hope this prev. behavior was a bug/undefined-behavior, since this CL changes that.

BUG=skia:5595
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2195893002

Review-Url: https://codereview.chromium.org/2195893002
2016-08-02 06:12:06 -07:00

374 lines
14 KiB
Python
Executable File

#
# Copyright 2015 Google Inc.
#
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.
#
#!/usr/bin/env python
usage = '''
Write extra flags to outfile for DM based on the bot name:
$ python dm_flags.py outfile Test-Ubuntu-GCC-GCE-CPU-AVX2-x86-Debug
Or run self-tests:
$ python dm_flags.py test
'''
import inspect
import json
import os
import sys
def lineno():
caller = inspect.stack()[1] # Up one level to our caller.
return inspect.getframeinfo(caller[0]).lineno
cov_start = lineno()+1 # We care about coverage starting just past this def.
def get_args(bot):
args = []
# 32-bit desktop bots tend to run out of memory, because they have relatively
# far more cores than RAM (e.g. 32 cores, 3G RAM). Hold them back a bit.
if '-x86-' in bot and not 'NexusPlayer' in bot:
args.extend('--threads 4'.split(' '))
# These are the canonical configs that we would ideally run on all bots. We
# may opt out or substitute some below for specific bots
configs = ['565', '8888', 'gpu', 'gpusrgb', 'pdf']
# Add in either msaa4 or msaa16 to the canonical set of configs to run
if 'Android' in bot or 'iOS' in bot:
configs.append('msaa4')
else:
configs.append('msaa16')
# With msaa, the S4 crashes and the NP produces a long error stream when we
# run with MSAA. The Tegra2 and Tegra3 just don't support it. No record of
# why we're not running msaa on iOS, probably started with gpu config and just
# haven't tried.
if ('GalaxyS4' in bot or
'NexusPlayer' in bot or
'Tegra3' in bot or
'iOS' in bot):
configs = [x for x in configs if 'msaa' not in x]
# Runs out of memory on Android bots and Daisy. Everyone else seems fine.
if 'Android' in bot or 'Daisy' in bot:
configs.remove('pdf')
if '-GCE-' in bot:
configs.extend(['f16', 'srgb']) # Gamma-correct formats.
configs.extend(['sp-8888', '2ndpic-8888']) # Test niche uses of SkPicture.
if '-TSAN' not in bot:
if ('TegraK1' in bot or
'TegraX1' in bot or
'GTX550Ti' in bot or
'GTX660' in bot or
'GT610' in bot):
if 'Android' in bot:
configs.append('nvprdit4')
else:
configs.append('nvprdit16')
# We want to test the OpenGL config not the GLES config on the X1
if 'TegraX1' in bot:
configs = [x.replace('gpu', 'gl') for x in configs]
configs = [x.replace('msaa', 'glmsaa') for x in configs]
configs = [x.replace('nvpr', 'glnvpr') for x in configs]
# NP is running out of RAM when we run all these modes. skia:3255
if 'NexusPlayer' not in bot:
configs.extend(mode + '-8888' for mode in
['serialize', 'tiles_rt', 'pic'])
if 'ANGLE' in bot:
configs.append('angle')
# We want to run gpudft on atleast the mali 400
if 'GalaxyS3' in bot:
configs.append('gpudft')
# Test instanced rendering on a limited number of platforms
if 'Nexus6' in bot:
configs.append('esinst') # esinst4 isn't working yet on Adreno.
elif 'TegraX1' in bot:
# Multisampled instanced configs use nvpr.
configs = [x.replace('glnvpr', 'glinst') for x in configs]
configs.append('glinst')
elif 'MacMini6.2' in bot:
configs.extend(['glinst', 'glinst16'])
# CommandBuffer bot *only* runs the command_buffer config.
if 'CommandBuffer' in bot:
configs = ['commandbuffer']
# Vulkan bot *only* runs the vk config.
if 'Vulkan' in bot:
configs = ['vk']
args.append('--config')
args.extend(configs)
# Run tests, gms, and image decoding tests everywhere.
args.extend('--src tests gm image'.split(' '))
if 'GalaxyS' in bot:
args.extend(('--threads', '0'))
blacklist = []
# TODO: ???
blacklist.extend('f16 _ _ dstreadshuffle'.split(' '))
blacklist.extend('f16 image _ _'.split(' '))
blacklist.extend('srgb image _ _'.split(' '))
blacklist.extend('gpusrgb image _ _'.split(' '))
# Certain gm's on win7 gpu and pdf are never finishing and keeping the test
# running forever
if 'Win7' in bot:
blacklist.extend('msaa16 gm _ colorwheelnative'.split(' '))
blacklist.extend('pdf gm _ fontmgr_iter_factory'.split(' '))
if 'Valgrind' in bot:
# These take 18+ hours to run.
blacklist.extend('pdf gm _ fontmgr_iter'.split(' '))
blacklist.extend('pdf _ _ PANO_20121023_214540.jpg'.split(' '))
blacklist.extend('pdf skp _ worldjournal'.split(' '))
blacklist.extend('pdf skp _ desk_baidu.skp'.split(' '))
blacklist.extend('pdf skp _ desk_wikipedia.skp'.split(' '))
if 'iOS' in bot:
blacklist.extend('gpu skp _ _ msaa skp _ _'.split(' '))
blacklist.extend('msaa16 gm _ tilemodesProcess'.split(' '))
if 'Mac' in bot or 'iOS' in bot:
# CG fails on questionable bmps
blacklist.extend('_ image gen_platf rgba32abf.bmp'.split(' '))
blacklist.extend('_ image gen_platf rgb24prof.bmp'.split(' '))
blacklist.extend('_ image gen_platf rgb24lprof.bmp'.split(' '))
blacklist.extend('_ image gen_platf 8bpp-pixeldata-cropped.bmp'.split(' '))
blacklist.extend('_ image gen_platf 4bpp-pixeldata-cropped.bmp'.split(' '))
blacklist.extend('_ image gen_platf 32bpp-pixeldata-cropped.bmp'.split(' '))
blacklist.extend('_ image gen_platf 24bpp-pixeldata-cropped.bmp'.split(' '))
# CG has unpredictable behavior on this questionable gif
# It's probably using uninitialized memory
blacklist.extend('_ image gen_platf frame_larger_than_image.gif'.split(' '))
# WIC fails on questionable bmps
if 'Win' in bot:
blacklist.extend('_ image gen_platf rle8-height-negative.bmp'.split(' '))
blacklist.extend('_ image gen_platf rle4-height-negative.bmp'.split(' '))
blacklist.extend('_ image gen_platf pal8os2v2.bmp'.split(' '))
blacklist.extend('_ image gen_platf pal8os2v2-16.bmp'.split(' '))
blacklist.extend('_ image gen_platf rgba32abf.bmp'.split(' '))
blacklist.extend('_ image gen_platf rgb24prof.bmp'.split(' '))
blacklist.extend('_ image gen_platf rgb24lprof.bmp'.split(' '))
blacklist.extend('_ image gen_platf 8bpp-pixeldata-cropped.bmp'.split(' '))
blacklist.extend('_ image gen_platf 4bpp-pixeldata-cropped.bmp'.split(' '))
blacklist.extend('_ image gen_platf 32bpp-pixeldata-cropped.bmp'.split(' '))
blacklist.extend('_ image gen_platf 24bpp-pixeldata-cropped.bmp'.split(' '))
if 'x86_64' in bot and 'CPU' in bot:
# This GM triggers a SkSmallAllocator assert.
blacklist.extend('_ gm _ composeshader_bitmap'.split(' '))
if 'Android' in bot or 'iOS' in bot:
# This test crashes the N9 (perhaps because of large malloc/frees). It also
# is fairly slow and not platform-specific. So we just disable it on all of
# Android and iOS. skia:5438
blacklist.extend('_ test _ GrShape'.split(' '))
if 'Win8' in bot:
# bungeman: "Doesn't work on Windows anyway, produces unstable GMs with
# 'Unexpected error' from DirectWrite"
blacklist.extend('_ gm _ fontscalerdistortable'.split(' '))
# skia:4095
bad_serialize_gms = ['bleed_image',
'c_gms',
'colortype',
'colortype_xfermodes',
'drawfilter',
'fontmgr_bounds_0.75_0',
'fontmgr_bounds_1_-0.25',
'fontmgr_bounds',
'fontmgr_match',
'fontmgr_iter']
# skia:5589
bad_serialize_gms.extend(['bitmapfilters',
'bitmapshaders',
'bleed',
'bleed_alpha_bmp',
'bleed_alpha_bmp_shader',
'convex_poly_clip',
'extractalpha',
'filterbitmap_checkerboard_32_32_g8',
'filterbitmap_image_mandrill_64',
'shadows',
'simpleaaclip_aaclip'])
# skia:5595
bad_serialize_gms.extend(['composeshader_bitmap',
'scaled_tilemodes_npot',
'scaled_tilemodes'])
for test in bad_serialize_gms:
blacklist.extend(['serialize-8888', 'gm', '_', test])
if 'Mac' not in bot:
for test in ['bleed_alpha_image', 'bleed_alpha_image_shader']:
blacklist.extend(['serialize-8888', 'gm', '_', test])
# It looks like we skip these only for out-of-memory concerns.
if 'Win' in bot or 'Android' in bot:
for test in ['verylargebitmap', 'verylarge_picture_image']:
blacklist.extend(['serialize-8888', 'gm', '_', test])
# skia:4769
for test in ['drawfilter']:
blacklist.extend([ 'sp-8888', 'gm', '_', test])
blacklist.extend([ 'pic-8888', 'gm', '_', test])
blacklist.extend(['2ndpic-8888', 'gm', '_', test])
# skia:4703
for test in ['image-cacherator-from-picture',
'image-cacherator-from-raster',
'image-cacherator-from-ctable']:
blacklist.extend([ 'sp-8888', 'gm', '_', test])
blacklist.extend([ 'pic-8888', 'gm', '_', test])
blacklist.extend([ '2ndpic-8888', 'gm', '_', test])
blacklist.extend(['serialize-8888', 'gm', '_', test])
# Extensions for RAW images
r = ["arw", "cr2", "dng", "nef", "nrw", "orf", "raf", "rw2", "pef", "srw",
"ARW", "CR2", "DNG", "NEF", "NRW", "ORF", "RAF", "RW2", "PEF", "SRW"]
# skbug.com/4888
# Blacklist RAW images (and a few large PNGs) on GPU bots
# until we can resolve failures
if 'GPU' in bot:
blacklist.extend('_ image _ interlaced1.png'.split(' '))
blacklist.extend('_ image _ interlaced2.png'.split(' '))
blacklist.extend('_ image _ interlaced3.png'.split(' '))
for raw_ext in r:
blacklist.extend(('_ image _ .%s' % raw_ext).split(' '))
if 'Nexus9' in bot:
for raw_ext in r:
blacklist.extend(('_ image _ .%s' % raw_ext).split(' '))
# Large image that overwhelms older Mac bots
if 'MacMini4.1-GPU' in bot:
blacklist.extend('_ image _ abnormal.wbmp'.split(' '))
blacklist.extend(['msaa16', 'gm', '_', 'blurcircles'])
match = []
if 'Valgrind' in bot: # skia:3021
match.append('~Threaded')
if 'GalaxyS3' in bot: # skia:1699
match.append('~WritePixels')
if 'AndroidOne' in bot: # skia:4711
match.append('~WritePixels')
if 'NexusPlayer' in bot:
match.append('~ResourceCache')
if 'Nexus10' in bot: # skia:5509
match.append('~CopySurface')
if 'GalaxyS4' in bot: # skia:4079
match.append('~imagefiltersclipped')
match.append('~imagefilterscropexpand')
match.append('~scaled_tilemodes_npot')
match.append('~bleed_image') # skia:4367
match.append('~ReadPixels') # skia:4368
if 'ANGLE' in bot and 'Debug' in bot:
match.append('~GLPrograms') # skia:4717
if 'MSAN' in bot:
match.extend(['~Once', '~Shared']) # Not sure what's up with these tests.
if 'TSAN' in bot:
match.extend(['~ReadWriteAlpha']) # Flaky on TSAN-covered on nvidia bots.
if blacklist:
args.append('--blacklist')
args.extend(blacklist)
if match:
args.append('--match')
args.extend(match)
# These bots run out of memory running RAW codec tests. Do not run them in
# parallel
if ('NexusPlayer' in bot or 'Nexus5' in bot or 'Nexus9' in bot
or 'Win8-MSVC-ShuttleB' in bot):
args.append('--noRAW_threading')
return args
cov_end = lineno() # Don't care about code coverage past here.
def self_test():
args = {}
cases = [
'Pretend-iOS-Bot',
'Test-Android-GCC-AndroidOne-GPU-Mali400MP2-Arm7-Release',
'Test-Android-GCC-GalaxyS3-GPU-Mali400-Arm7-Debug',
'Test-Android-GCC-GalaxyS4-GPU-SGX544-Arm7-Release',
'Test-Android-GCC-Nexus6-GPU-Adreno420-Arm7-Debug',
'Test-Android-GCC-Nexus7-GPU-Tegra3-Arm7-Release',
'Test-Android-GCC-Nexus9-GPU-TegraK1-Arm64-Debug',
'Test-Android-GCC-Nexus10-GPU-MaliT604-Arm7-Debug',
'Test-Android-GCC-NexusPlayer-CPU-SSSE3-x86-Release',
'Test-Android-GCC-NVIDIA_Shield-GPU-TegraX1-Arm64-Release',
'Test-Mac-Clang-MacMini4.1-GPU-GeForce320M-x86_64-Release',
'Test-Mac-Clang-MacMini6.2-GPU-HD4000-x86_64-Debug-CommandBuffer',
'Test-Mac-Clang-MacMini6.2-GPU-HD4000-x86_64-Debug',
'Test-Mac10.8-Clang-MacMini4.1-CPU-SSE4-x86_64-Release',
'Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Debug-MSAN',
'Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-TSAN',
'Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Valgrind',
'Test-Ubuntu-GCC-ShuttleA-GPU-GTX550Ti-x86_64-Release-Valgrind',
'Test-Win-MSVC-GCE-CPU-AVX2-x86_64-Debug',
'Test-Win10-MSVC-ShuttleA-GPU-GTX660-x86_64-Debug-Vulkan',
'Test-Win7-MSVC-ShuttleA-GPU-HD2000-x86-Debug-ANGLE',
'Test-Win8-MSVC-GCE-CPU-AVX2-x86_64-Debug',
]
this_file = os.path.basename(__file__)
try:
import coverage
cov = coverage.coverage()
cov.start()
for case in cases:
args[case] = get_args(case)
cov.stop()
_, _, not_run, _ = cov.analysis(this_file)
filtered = [line for line in not_run if line > cov_start and line < cov_end]
if filtered:
print 'Lines not covered by test cases: ', filtered
sys.exit(1)
except ImportError:
print ("We cannot guarantee that this files tests are comprehensive " +
"without coverage.py. Please install it when you get a chance.")
golden = this_file.replace('.py', '.json')
with open(os.path.join(os.path.dirname(__file__), golden), 'w') as f:
json.dump(args, f, indent=2, sort_keys=True)
if __name__ == '__main__':
if len(sys.argv) == 2 and sys.argv[1] == 'test':
self_test()
sys.exit(0)
if len(sys.argv) != 3:
print usage
sys.exit(1)
with open(sys.argv[1], 'w') as out:
json.dump(get_args(sys.argv[2]), out)