skia2/tools/rebaseline.py
2013-07-17 00:09:10 +00:00

371 lines
16 KiB
Python
Executable File

#!/usr/bin/python
'''
Copyright 2012 Google Inc.
Use of this source code is governed by a BSD-style license that can be
found in the LICENSE file.
'''
'''
Rebaselines the given GM tests, on all bots and all configurations.
Must be run from the gm-expected directory. If run from a git or SVN
checkout, the files will be added to the staging area for commit.
TODO(epoger): Fix indentation in this file (2-space indents, not 4-space).
'''
# System-level imports
import argparse
import os
import re
import sys
import urllib2
# Imports from local directory
import rebaseline_imagefiles
# Imports from within Skia
#
# We need to add the 'gm' directory, so that we can import gm_json.py within
# that directory. That script allows us to parse the actual-results.json file
# written out by the GM tool.
# Make sure that the 'gm' dir is in the PYTHONPATH, but add it at the *end*
# so any dirs that are already in the PYTHONPATH will be preferred.
#
# This assumes that the 'gm' directory has been checked out as a sibling of
# the 'tools' directory containing this script, which will be the case if
# 'trunk' was checked out as a single unit.
GM_DIRECTORY = os.path.realpath(
os.path.join(os.path.dirname(os.path.dirname(__file__)), 'gm'))
if GM_DIRECTORY not in sys.path:
sys.path.append(GM_DIRECTORY)
import gm_json
# Mapping of gm-expectations subdir (under
# https://skia.googlecode.com/svn/gm-expected/ )
# to builder name (see list at http://108.170.217.252:10117/builders )
SUBDIR_MAPPING = {
'base-shuttle-win7-intel-float':
'Test-Win7-ShuttleA-HD2000-x86-Release',
'base-shuttle-win7-intel-angle':
'Test-Win7-ShuttleA-HD2000-x86-Release-ANGLE',
'base-shuttle-win7-intel-directwrite':
'Test-Win7-ShuttleA-HD2000-x86-Release-DirectWrite',
'base-shuttle_ubuntu12_ati5770':
'Test-Ubuntu12-ShuttleA-ATI5770-x86_64-Release',
'base-macmini':
'Test-Mac10.6-MacMini4.1-GeForce320M-x86-Release',
'base-macmini-lion-float':
'Test-Mac10.7-MacMini4.1-GeForce320M-x86-Release',
'base-android-galaxy-nexus':
'Test-Android-GalaxyNexus-SGX540-Arm7-Debug',
'base-android-nexus-7':
'Test-Android-Nexus7-Tegra3-Arm7-Release',
'base-android-nexus-s':
'Test-Android-NexusS-SGX540-Arm7-Release',
'base-android-xoom':
'Test-Android-Xoom-Tegra2-Arm7-Release',
'base-android-nexus-10':
'Test-Android-Nexus10-MaliT604-Arm7-Release',
'base-android-nexus-4':
'Test-Android-Nexus4-Adreno320-Arm7-Release',
}
class _InternalException(Exception):
pass
# Object that handles exceptions, either raising them immediately or collecting
# them to display later on.
class ExceptionHandler(object):
# params:
# keep_going_on_failure: if False, report failures and quit right away;
# if True, collect failures until
# ReportAllFailures() is called
def __init__(self, keep_going_on_failure=False):
self._keep_going_on_failure = keep_going_on_failure
self._failures_encountered = []
self._exiting = False
# Exit the program with the given status value.
def _Exit(self, status=1):
self._exiting = True
sys.exit(status)
# We have encountered an exception; either collect the info and keep going,
# or exit the program right away.
def RaiseExceptionOrContinue(self, e):
# If we are already quitting the program, propagate any exceptions
# so that the proper exit status will be communicated to the shell.
if self._exiting:
raise e
if self._keep_going_on_failure:
print >> sys.stderr, 'WARNING: swallowing exception %s' % e
self._failures_encountered.append(e)
else:
print >> sys.stderr, e
print >> sys.stderr, (
'Halting at first exception; to keep going, re-run ' +
'with the --keep-going-on-failure option set.')
self._Exit()
def ReportAllFailures(self):
if self._failures_encountered:
print >> sys.stderr, ('Encountered %d failures (see above).' %
len(self._failures_encountered))
self._Exit()
# Object that rebaselines a JSON expectations file (not individual image files).
class JsonRebaseliner(object):
# params:
# expectations_root: root directory of all expectations JSON files
# expectations_filename: filename (under expectations_root) of JSON
# expectations file; typically
# "expected-results.json"
# actuals_base_url: base URL from which to read actual-result JSON files
# actuals_filename: filename (under actuals_base_url) from which to read a
# summary of results; typically "actual-results.json"
# exception_handler: reference to rebaseline.ExceptionHandler object
# tests: list of tests to rebaseline, or None if we should rebaseline
# whatever files the JSON results summary file tells us to
# configs: which configs to run for each test, or None if we should
# rebaseline whatever configs the JSON results summary file tells
# us to
# add_new: if True, add expectations for tests which don't have any yet
def __init__(self, expectations_root, expectations_filename,
actuals_base_url, actuals_filename, exception_handler,
tests=None, configs=None, add_new=False):
self._expectations_root = expectations_root
self._expectations_filename = expectations_filename
self._tests = tests
self._configs = configs
self._actuals_base_url = actuals_base_url
self._actuals_filename = actuals_filename
self._exception_handler = exception_handler
self._add_new = add_new
self._image_filename_re = re.compile(gm_json.IMAGE_FILENAME_PATTERN)
# Returns the full contents of filepath, as a single string.
# If filepath looks like a URL, try to read it that way instead of as
# a path on local storage.
#
# Raises _InternalException if there is a problem.
def _GetFileContents(self, filepath):
if filepath.startswith('http:') or filepath.startswith('https:'):
try:
return urllib2.urlopen(filepath).read()
except urllib2.HTTPError as e:
raise _InternalException('unable to read URL %s: %s' % (
filepath, e))
else:
return open(filepath, 'r').read()
# Returns a dictionary of actual results from actual-results.json file.
#
# The dictionary returned has this format:
# {
# u'imageblur_565.png': [u'bitmap-64bitMD5', 3359963596899141322],
# u'imageblur_8888.png': [u'bitmap-64bitMD5', 4217923806027861152],
# u'shadertext3_8888.png': [u'bitmap-64bitMD5', 3713708307125704716]
# }
#
# If the JSON actual result summary file cannot be loaded, logs a warning
# message and returns None.
# If the JSON actual result summary file can be loaded, but we have
# trouble parsing it, raises an Exception.
#
# params:
# json_url: URL pointing to a JSON actual result summary file
# sections: a list of section names to include in the results, e.g.
# [gm_json.JSONKEY_ACTUALRESULTS_FAILED,
# gm_json.JSONKEY_ACTUALRESULTS_NOCOMPARISON] ;
# if None, then include ALL sections.
def _GetActualResults(self, json_url, sections=None):
try:
json_contents = self._GetFileContents(json_url)
except _InternalException:
print >> sys.stderr, (
'could not read json_url %s ; skipping this platform.' %
json_url)
return None
json_dict = gm_json.LoadFromString(json_contents)
results_to_return = {}
actual_results = json_dict[gm_json.JSONKEY_ACTUALRESULTS]
if not sections:
sections = actual_results.keys()
for section in sections:
section_results = actual_results[section]
if section_results:
results_to_return.update(section_results)
return results_to_return
# Rebaseline all tests/types we specified in the constructor,
# within this gm-expectations subdir.
#
# params:
# subdir : e.g. 'base-shuttle-win7-intel-float'
# builder : e.g. 'Test-Win7-ShuttleA-HD2000-x86-Release'
def RebaselineSubdir(self, subdir, builder):
# Read in the actual result summary, and extract all the tests whose
# results we need to update.
actuals_url = '/'.join([self._actuals_base_url,
subdir, builder, subdir,
self._actuals_filename])
# In most cases, we won't need to re-record results that are already
# succeeding, but including the SUCCEEDED results will allow us to
# re-record expectations if they somehow get out of sync.
sections = [gm_json.JSONKEY_ACTUALRESULTS_FAILED,
gm_json.JSONKEY_ACTUALRESULTS_SUCCEEDED]
if self._add_new:
sections.append(gm_json.JSONKEY_ACTUALRESULTS_NOCOMPARISON)
results_to_update = self._GetActualResults(json_url=actuals_url,
sections=sections)
# Read in current expectations.
expectations_json_filepath = os.path.join(
self._expectations_root, subdir, self._expectations_filename)
expectations_dict = gm_json.LoadFromFile(expectations_json_filepath)
expected_results = expectations_dict[gm_json.JSONKEY_EXPECTEDRESULTS]
# Update the expectations in memory, skipping any tests/configs that
# the caller asked to exclude.
skipped_images = []
if results_to_update:
for (image_name, image_results) in results_to_update.iteritems():
(test, config) = \
self._image_filename_re.match(image_name).groups()
if self._tests:
if test not in self._tests:
skipped_images.append(image_name)
continue
if self._configs:
if config not in self._configs:
skipped_images.append(image_name)
continue
if not expected_results.get(image_name):
expected_results[image_name] = {}
expected_results[image_name] \
[gm_json.JSONKEY_EXPECTEDRESULTS_ALLOWEDDIGESTS] = \
[image_results]
# Write out updated expectations.
gm_json.WriteToFile(expectations_dict, expectations_json_filepath)
# main...
parser = argparse.ArgumentParser()
parser.add_argument('--actuals-base-url',
help='base URL from which to read files containing JSON ' +
'summaries of actual GM results; defaults to %(default)s',
default='http://skia-autogen.googlecode.com/svn/gm-actual')
parser.add_argument('--actuals-filename',
help='filename (within platform-specific subdirectories ' +
'of ACTUALS_BASE_URL) to read a summary of results from; ' +
'defaults to %(default)s',
default='actual-results.json')
# TODO(epoger): Add test that exercises --add-new argument.
parser.add_argument('--add-new', action='store_true',
help='in addition to the standard behavior of ' +
'updating expectations for failing tests, add ' +
'expectations for tests which don\'t have expectations ' +
'yet.')
# TODO(epoger): Add test that exercises --configs argument.
parser.add_argument('--configs', metavar='CONFIG', nargs='+',
help='which configurations to rebaseline, e.g. ' +
'"--configs 565 8888", as a filter over the full set of ' +
'results in ACTUALS_FILENAME; if unspecified, rebaseline ' +
'*all* configs that are available.')
# TODO(epoger): The --dry-run argument will no longer be needed once we
# are only rebaselining JSON files.
parser.add_argument('--dry-run', action='store_true',
help='instead of actually downloading files or adding ' +
'files to checkout, display a list of operations that ' +
'we would normally perform')
parser.add_argument('--expectations-filename',
help='filename (under EXPECTATIONS_ROOT) to read ' +
'current expectations from, and to write new ' +
'expectations into; defaults to %(default)s',
default='expected-results.json')
parser.add_argument('--expectations-root',
help='root of expectations directory to update-- should ' +
'contain one or more base-* subdirectories. Defaults to ' +
'%(default)s',
default='.')
parser.add_argument('--keep-going-on-failure', action='store_true',
help='instead of halting at the first error encountered, ' +
'keep going and rebaseline as many tests as possible, ' +
'and then report the full set of errors at the end')
parser.add_argument('--subdirs', metavar='SUBDIR', nargs='+',
help='which platform subdirectories to rebaseline; ' +
'if unspecified, rebaseline all subdirs, same as ' +
'"--subdirs %s"' % ' '.join(sorted(SUBDIR_MAPPING.keys())))
# TODO(epoger): Add test that exercises --tests argument.
parser.add_argument('--tests', metavar='TEST', nargs='+',
help='which tests to rebaseline, e.g. ' +
'"--tests aaclip bigmatrix", as a filter over the full ' +
'set of results in ACTUALS_FILENAME; if unspecified, ' +
'rebaseline *all* tests that are available.')
args = parser.parse_args()
exception_handler = ExceptionHandler(
keep_going_on_failure=args.keep_going_on_failure)
if args.subdirs:
subdirs = args.subdirs
missing_json_is_fatal = True
else:
subdirs = sorted(SUBDIR_MAPPING.keys())
missing_json_is_fatal = False
for subdir in subdirs:
if not subdir in SUBDIR_MAPPING.keys():
raise Exception(('unrecognized platform subdir "%s"; ' +
'should be one of %s') % (
subdir, SUBDIR_MAPPING.keys()))
builder = SUBDIR_MAPPING[subdir]
# We instantiate different Rebaseliner objects depending
# on whether we are rebaselining an expected-results.json file, or
# individual image files. Different gm-expected subdirectories may move
# from individual image files to JSON-format expectations at different
# times, so we need to make this determination per subdirectory.
#
# See https://goto.google.com/ChecksumTransitionDetail
expectations_json_file = os.path.join(args.expectations_root, subdir,
args.expectations_filename)
if os.path.isfile(expectations_json_file):
rebaseliner = JsonRebaseliner(
expectations_root=args.expectations_root,
expectations_filename=args.expectations_filename,
tests=args.tests, configs=args.configs,
actuals_base_url=args.actuals_base_url,
actuals_filename=args.actuals_filename,
exception_handler=exception_handler,
add_new=args.add_new)
else:
# TODO(epoger): When we get rid of the ImageRebaseliner implementation,
# we should raise an Exception in this case (no JSON expectations file
# found to update), to prevent a recurrence of
# https://code.google.com/p/skia/issues/detail?id=1403 ('rebaseline.py
# script fails with misleading output when run outside of gm-expected
# dir')
rebaseliner = rebaseline_imagefiles.ImageRebaseliner(
expectations_root=args.expectations_root,
tests=args.tests, configs=args.configs,
dry_run=args.dry_run,
json_base_url=args.actuals_base_url,
json_filename=args.actuals_filename,
exception_handler=exception_handler,
add_new=args.add_new,
missing_json_is_fatal=missing_json_is_fatal)
try:
rebaseliner.RebaselineSubdir(subdir=subdir, builder=builder)
except BaseException as e:
exception_handler.RaiseExceptionOrContinue(e)
exception_handler.ReportAllFailures()