def __init__(self,
python_setup,
python_repos,
ivy_bootstrapper,
thrift_binary_factory,
interpreter,
builder,
targets,
platforms,
extra_requirements=None,
log=None):
self._python_setup = python_setup
self._python_repos = python_repos
self._ivy_bootstrapper = ivy_bootstrapper
self._thrift_binary_factory = thrift_binary_factory
self._interpreter = interpreter
self._builder = builder
self._targets = targets
self._platforms = platforms
self._extra_requirements = list(
extra_requirements) if extra_requirements else []
self._logger = log or logger
# Note: unrelated to the general pants artifact cache.
self._artifact_cache_root = os.path.join(
self._python_setup.artifact_cache_dir,
str(self._interpreter.identity))
self._key_generator = CacheKeyGenerator()
self._build_invalidator = BuildInvalidator(self._artifact_cache_root)
def __init__(self, context, workdir):
"""Subclass __init__ methods, if defined, *must* follow this idiom:
class MyTask(Task):
def __init__(self, *args, **kwargs):
super(MyTask, self).__init__(*args, **kwargs)
...
This allows us to change Task.__init__()'s arguments without
changing every subclass. If the subclass does not need its own
initialization, this method can (and should) be omitted entirely.
"""
super(TaskBase, self).__init__()
self.context = context
self._workdir = workdir
# TODO: It would be nice to use self.get_options().cache_key_gen_version here, because then
# we could have a separate value for each scope if we really wanted to. However we can't
# access per-task options in Task.__init__ because GroupTask.__init__ calls it with the
# group task's scope, which isn't currently in the known scopes we generate options for.
self._cache_key_generator = CacheKeyGenerator(
self.context.options.for_global_scope().cache_key_gen_version)
self._cache_key_errors = set()
self._build_invalidator_dir = os.path.join(
self.context.options.for_global_scope().pants_workdir,
'build_invalidator', self.stable_name())
self._cache_factory = CacheSetup.create_cache_factory_for_task(self)
self._options_fingerprinter = OptionsFingerprinter(
self.context.build_graph)
self._fingerprint = None
def _do_invalidation_check(self,
fingerprint_strategy,
invalidate_dependents,
targets,
topological_order):
if self._cache_factory.ignore:
cache_key_generator = UncacheableCacheKeyGenerator()
else:
cache_key_generator = CacheKeyGenerator(
self.context.options.for_global_scope().cache_key_gen_version,
self.fingerprint)
cache_manager = InvalidationCacheManager(self.workdir,
cache_key_generator,
self._build_invalidator(),
invalidate_dependents,
fingerprint_strategy=fingerprint_strategy,
invalidation_report=self.context.invalidation_report,
task_name=self._task_name,
task_version=self.implementation_version_str(),
artifact_write_callback=self.maybe_write_artifact)
# If this Task's execution has been forced, invalidate all our target fingerprints.
if self._cache_factory.ignore and not self._force_invalidated:
self.invalidate()
self._force_invalidated = True
return cache_manager.check(targets, topological_order=topological_order)
def __init__(self, cache_manager, versioned_targets):
self._cache_manager = cache_manager
self.versioned_targets = versioned_targets
self.targets = [vt.target for vt in versioned_targets]
# The following line is a no-op if cache_key was set in the VersionedTarget __init__ method.
self.cache_key = CacheKeyGenerator.combine_cache_keys(
[vt.cache_key for vt in versioned_targets])
# NB: previous_cache_key may be None on the first build of a target.
self.previous_cache_key = cache_manager.previous_key(self.cache_key)
self.valid = self.previous_cache_key == self.cache_key
if cache_manager.invalidation_report:
cache_manager.invalidation_report.add_vts(cache_manager,
self.targets,
self.cache_key,
self.valid,
phase='init')
self._results_dir = None
self._current_results_dir = None
self._previous_results_dir = None
# True if the results_dir for this VT was created incrementally via clone of the
# previous results_dir.
self.is_incremental = False
def __init__(self,
python_setup,
python_repos,
ivy_bootstrapper,
thrift_binary_factory,
interpreter,
builder,
targets,
platforms,
extra_requirements=None,
log=None):
self._python_setup = python_setup
self._python_repos = python_repos
self._ivy_bootstrapper = ivy_bootstrapper
self._thrift_binary_factory = thrift_binary_factory
self._interpreter = interpreter
self._builder = builder
self._targets = targets
self._platforms = platforms
self._extra_requirements = list(extra_requirements) if extra_requirements else []
self._logger = log or logger
# Note: unrelated to the general pants artifact cache.
self._artifact_cache_root = os.path.join(
self._python_setup.artifact_cache_dir, str(self._interpreter.identity))
self._key_generator = CacheKeyGenerator()
self._build_invalidator = BuildInvalidator(self._artifact_cache_root)
def __init__(self, cache_manager, versioned_targets):
"""
:API: public
"""
self._cache_manager = cache_manager
self.versioned_targets = versioned_targets
self.targets = [vt.target for vt in versioned_targets]
# The following line is a no-op if cache_key was set in the VersionedTarget __init__ method.
self.cache_key = CacheKeyGenerator.combine_cache_keys([vt.cache_key for vt in versioned_targets])
# NB: previous_cache_key may be None on the first build of a target.
self.previous_cache_key = cache_manager.previous_key(self.cache_key)
self.valid = self.previous_cache_key == self.cache_key
self.num_chunking_units = self.cache_key.num_chunking_units
if cache_manager.invalidation_report:
cache_manager.invalidation_report.add_vts(
cache_manager, self.targets, self.cache_key, self.valid, phase="init"
)
self._results_dir = None
self._previous_results_dir = None
# True if the results_dir for this VT was created incrementally via clone of the
# previous results_dir.
self.is_incremental = False
def setUp(self):
super(InvalidationCacheManagerTest, self).setUp()
self._dir = tempfile.mkdtemp()
self.cache_manager = InvalidationCacheManager(
results_dir_root=os.path.join(self._dir, 'results'),
cache_key_generator=CacheKeyGenerator(),
build_invalidator=BuildInvalidator(os.path.join(self._dir, 'build_invalidator')),
invalidate_dependents=True,
)
def setUp(self):
super().setUp()
self._dir = tempfile.mkdtemp()
self.cache_manager = InvalidationCacheManager(
results_dir_root=os.path.join(self._dir, "results"),
cache_key_generator=CacheKeyGenerator(),
build_invalidator=BuildInvalidator(
os.path.join(self._dir, "build_invalidator")),
invalidate_dependents=True,
task_version_slug="deadbeef",
)
def __init__(self, cache_manager, versioned_targets):
self._cache_manager = cache_manager
self.versioned_targets = versioned_targets
self.targets = [vt.target for vt in versioned_targets]
# The following line is a no-op if cache_key was set in the VersionedTarget __init__ method.
self.cache_key = CacheKeyGenerator.combine_cache_keys([vt.cache_key for vt in versioned_targets])
self.num_chunking_units = self.cache_key.num_chunking_units
self.valid = not cache_manager.needs_update(self.cache_key)
if cache_manager.invalidation_report:
cache_manager.invalidation_report.add_vts(
cache_manager, self.targets, self.cache_key, self.valid, phase="init"
)
def __init__(self, cache_manager, versioned_targets):
self._cache_manager = cache_manager
self.versioned_targets = versioned_targets
self.targets = [vt.target for vt in versioned_targets]
# The following line is a no-op if cache_key was set in the VersionedTarget __init__ method.
self.cache_key = CacheKeyGenerator.combine_cache_keys(
[vt.cache_key for vt in versioned_targets])
self.num_chunking_units = self.cache_key.num_chunking_units
self.valid = not cache_manager.needs_update(self.cache_key)
if cache_manager.invalidation_report:
cache_manager.invalidation_report.add_vts(cache_manager,
self.targets,
self.cache_key,
self.valid,
phase='init')
class PythonChroot(object):
_VALID_DEPENDENCIES = {
Files: 'files',
PrepCommand: 'prep',
PythonLibrary: 'libraries',
PythonRequirementLibrary: 'reqs',
PythonBinary: 'binaries',
PythonThriftLibrary: 'thrifts',
PythonAntlrLibrary: 'antlrs',
PythonTests: 'tests',
Resources: 'resources'
}
class InvalidDependencyException(Exception):
def __init__(self, target):
super(PythonChroot.InvalidDependencyException, self).__init__(
'Not a valid Python dependency! Found: {}'.format(target))
@staticmethod
def get_platforms(platform_list):
return tuple({Platform.current() if p == 'current' else p for p in platform_list})
def __init__(self,
python_setup,
python_repos,
ivy_bootstrapper,
thrift_binary_factory,
interpreter,
builder,
targets,
platforms,
extra_requirements=None,
log=None):
self._python_setup = python_setup
self._python_repos = python_repos
self._ivy_bootstrapper = ivy_bootstrapper
self._thrift_binary_factory = thrift_binary_factory
self._interpreter = interpreter
self._builder = builder
self._targets = targets
self._platforms = platforms
self._extra_requirements = list(extra_requirements) if extra_requirements else []
self._logger = log or logger
# Note: unrelated to the general pants artifact cache.
self._artifact_cache_root = os.path.join(
self._python_setup.artifact_cache_dir, str(self._interpreter.identity))
self._key_generator = CacheKeyGenerator()
self._build_invalidator = BuildInvalidator(self._artifact_cache_root)
def delete(self):
"""Deletes this chroot from disk if it has been dumped."""
safe_rmtree(self.path())
def debug(self, msg):
self._logger.debug(msg)
def path(self):
return os.path.realpath(self._builder.path())
def pex(self):
return PEX(self.path(), interpreter=self._interpreter)
def package_pex(self, filename):
"""Package into a PEX zipfile.
:param filename: The filename where the PEX should be stored.
"""
self._builder.build(filename)
def _dump_library(self, library):
def copy_to_chroot(base, path, add_function):
src = os.path.join(get_buildroot(), base, path)
add_function(src, path)
self.debug(' Dumping library: {}'.format(library))
for relpath in library.sources_relative_to_source_root():
try:
copy_to_chroot(library.target_base, relpath, self._builder.add_source)
except OSError:
logger.error("Failed to copy {path} for library {library}"
.format(path=os.path.join(library.target_base, relpath),
library=library))
raise
for resources_tgt in library.resources:
for resource_file_from_source_root in resources_tgt.sources_relative_to_source_root():
try:
copy_to_chroot(resources_tgt.target_base, resource_file_from_source_root,
self._builder.add_resource)
except OSError:
logger.error("Failed to copy {path} for resource {resource}"
.format(path=os.path.join(resources_tgt.target_base,
resource_file_from_source_root),
resource=resources_tgt.address.spec))
raise
def _dump_requirement(self, req):
self.debug(' Dumping requirement: {}'.format(req))
self._builder.add_requirement(req)
def _dump_distribution(self, dist):
self.debug(' Dumping distribution: .../{}'.format(os.path.basename(dist.location)))
self._builder.add_distribution(dist)
def _generate_requirement(self, library, builder_cls):
library_key = self._key_generator.key_for_target(library)
builder = builder_cls(target=library,
root_dir=get_buildroot(),
target_suffix='-' + library_key.hash[:8])
cache_dir = os.path.join(self._artifact_cache_root, library_key.id)
if self._build_invalidator.needs_update(library_key):
sdist = builder.build(interpreter=self._interpreter)
safe_mkdir(cache_dir)
shutil.copy(sdist, os.path.join(cache_dir, os.path.basename(sdist)))
self._build_invalidator.update(library_key)
return PythonRequirement(builder.requirement_string(), repository=cache_dir, use_2to3=True)
def _generate_thrift_requirement(self, library):
thrift_builder = functools.partial(PythonThriftBuilder,
thrift_binary_factory=self._thrift_binary_factory,
workdir=safe_mkdtemp(dir=self.path(), prefix='thrift.'))
return self._generate_requirement(library, thrift_builder)
def _generate_antlr_requirement(self, library):
antlr_builder = functools.partial(PythonAntlrBuilder,
ivy_bootstrapper=self._ivy_bootstrapper,
workdir=safe_mkdtemp(dir=self.path(), prefix='antlr.'))
return self._generate_requirement(library, antlr_builder)
def resolve(self, targets):
children = defaultdict(OrderedSet)
def add_dep(trg):
for target_type, target_key in self._VALID_DEPENDENCIES.items():
if isinstance(trg, target_type):
children[target_key].add(trg)
return
elif type(trg) == Target:
return
raise self.InvalidDependencyException(trg)
for target in targets:
target.walk(add_dep)
return children
def dump(self):
self.debug('Building chroot for {}:'.format(self._targets))
targets = self.resolve(self._targets)
for lib in targets['libraries'] | targets['binaries']:
self._dump_library(lib)
generated_reqs = OrderedSet()
if targets['thrifts']:
for thr in targets['thrifts']:
generated_reqs.add(self._generate_thrift_requirement(thr))
generated_reqs.add(PythonRequirement('thrift', use_2to3=True))
for antlr in targets['antlrs']:
generated_reqs.add(self._generate_antlr_requirement(antlr))
reqs_from_libraries = OrderedSet()
for req_lib in targets['reqs']:
for req in req_lib.payload.requirements:
reqs_from_libraries.add(req)
reqs_to_build = OrderedSet()
find_links = OrderedSet()
for req in reqs_from_libraries | generated_reqs | self._extra_requirements:
if not req.should_build(self._interpreter.python, Platform.current()):
self.debug('Skipping {} based upon version filter'.format(req))
continue
reqs_to_build.add(req)
self._dump_requirement(req.requirement)
if req.repository:
find_links.add(req.repository)
distributions = self._resolve_multi(reqs_to_build, find_links)
locations = set()
for platform, dist_set in distributions.items():
for dist in dist_set:
if dist.location not in locations:
self._dump_distribution(dist)
locations.add(dist.location)
if len(targets['binaries']) > 1:
print('WARNING: Target has multiple python_binary targets!', file=sys.stderr)
return self._builder
def _resolve_multi(self, requirements, find_links):
"""Multi-platform dependency resolution for PEX files.
Given a pants configuration and a set of requirements, return a map of platform name -> list
of :class:`pkg_resources.Distribution` instances needed to satisfy them on that platform.
That may involve distributions for multiple platforms.
:param requirements: A list of :class:`PythonRequirement` objects to resolve.
:param find_links: Additional paths to search for source packages during resolution.
"""
distributions = dict()
platforms = self.get_platforms(self._platforms or self._python_setup.platforms)
fetchers = self._python_repos.get_fetchers()
fetchers.extend(Fetcher([path]) for path in find_links)
context = self._python_repos.get_network_context()
for platform in platforms:
requirements_cache_dir = os.path.join(self._python_setup.resolver_cache_dir,
str(self._interpreter.identity))
distributions[platform] = resolve(
requirements=[req.requirement for req in requirements],
interpreter=self._interpreter,
fetchers=fetchers,
platform=platform,
context=context,
cache=requirements_cache_dir,
cache_ttl=self._python_setup.resolver_cache_ttl,
allow_prereleases=self._python_setup.resolver_allow_prereleases)
return distributions
class PythonChroot(object):
_VALID_DEPENDENCIES = {
PrepCommand: 'prep',
PythonLibrary: 'libraries',
PythonRequirementLibrary: 'reqs',
PythonBinary: 'binaries',
PythonThriftLibrary: 'thrifts',
PythonAntlrLibrary: 'antlrs',
PythonTests: 'tests',
Resources: 'resources'
}
class InvalidDependencyException(Exception):
def __init__(self, target):
super(PythonChroot.InvalidDependencyException, self).__init__(
'Not a valid Python dependency! Found: {}'.format(target))
@staticmethod
def get_platforms(platform_list):
return tuple({
Platform.current() if p == 'current' else p
for p in platform_list
})
def __init__(self,
python_setup,
python_repos,
ivy_bootstrapper,
thrift_binary_factory,
interpreter,
builder,
targets,
platforms,
extra_requirements=None,
log=None):
self._python_setup = python_setup
self._python_repos = python_repos
self._ivy_bootstrapper = ivy_bootstrapper
self._thrift_binary_factory = thrift_binary_factory
self._interpreter = interpreter
self._builder = builder
self._targets = targets
self._platforms = platforms
self._extra_requirements = list(
extra_requirements) if extra_requirements else []
self._logger = log or logger
# Note: unrelated to the general pants artifact cache.
self._artifact_cache_root = os.path.join(
self._python_setup.artifact_cache_dir,
str(self._interpreter.identity))
self._key_generator = CacheKeyGenerator()
self._build_invalidator = BuildInvalidator(self._artifact_cache_root)
def delete(self):
"""Deletes this chroot from disk if it has been dumped."""
safe_rmtree(self.path())
def debug(self, msg):
self._logger.debug(msg)
def path(self):
return os.path.realpath(self._builder.path())
def pex(self):
return PEX(self.path(), interpreter=self._interpreter)
def package_pex(self, filename):
"""Package into a PEX zipfile.
:param filename: The filename where the PEX should be stored.
"""
self._builder.build(filename)
def _dump_library(self, library):
def copy_to_chroot(base, path, add_function):
src = os.path.join(get_buildroot(), base, path)
add_function(src, path)
self.debug(' Dumping library: {}'.format(library))
for relpath in library.sources_relative_to_source_root():
try:
copy_to_chroot(library.target_base, relpath,
self._builder.add_source)
except OSError:
logger.error(
"Failed to copy {path} for library {library}".format(
path=os.path.join(library.target_base, relpath),
library=library))
raise
for resources_tgt in library.resources:
for resource_file_from_source_root in resources_tgt.sources_relative_to_source_root(
):
try:
copy_to_chroot(resources_tgt.target_base,
resource_file_from_source_root,
self._builder.add_resource)
except OSError:
logger.error(
"Failed to copy {path} for resource {resource}".format(
path=os.path.join(resources_tgt.target_base,
resource_file_from_source_root),
resource=resources_tgt.address.spec))
raise
def _dump_requirement(self, req):
self.debug(' Dumping requirement: {}'.format(req))
self._builder.add_requirement(req)
def _dump_distribution(self, dist):
self.debug(' Dumping distribution: .../{}'.format(
os.path.basename(dist.location)))
self._builder.add_distribution(dist)
def _generate_requirement(self, library, builder_cls):
library_key = self._key_generator.key_for_target(library)
builder = builder_cls(target=library,
root_dir=get_buildroot(),
target_suffix='-' + library_key.hash[:8])
cache_dir = os.path.join(self._artifact_cache_root, library_key.id)
if self._build_invalidator.needs_update(library_key):
sdist = builder.build(interpreter=self._interpreter)
safe_mkdir(cache_dir)
shutil.copy(sdist, os.path.join(cache_dir,
os.path.basename(sdist)))
self._build_invalidator.update(library_key)
return PythonRequirement(builder.requirement_string(),
repository=cache_dir,
use_2to3=True)
def _generate_thrift_requirement(self, library):
thrift_builder = functools.partial(
PythonThriftBuilder,
thrift_binary_factory=self._thrift_binary_factory,
workdir=safe_mkdtemp(dir=self.path(), prefix='thrift.'))
return self._generate_requirement(library, thrift_builder)
def _generate_antlr_requirement(self, library):
antlr_builder = functools.partial(
PythonAntlrBuilder,
ivy_bootstrapper=self._ivy_bootstrapper,
workdir=safe_mkdtemp(dir=self.path(), prefix='antlr.'))
return self._generate_requirement(library, antlr_builder)
def resolve(self, targets):
children = defaultdict(OrderedSet)
def add_dep(trg):
# Currently we handle all of our code generation, so we don't want to operate over any
# synthetic targets injected upstream.
# TODO(John Sirois): Revisit this when building a proper python product pipeline.
if trg.is_synthetic:
return
for target_type, target_key in self._VALID_DEPENDENCIES.items():
if isinstance(trg, target_type):
children[target_key].add(trg)
return
elif type(trg) == Target:
return
raise self.InvalidDependencyException(trg)
for target in targets:
target.walk(add_dep)
return children
def dump(self):
self.debug('Building chroot for {}:'.format(self._targets))
targets = self.resolve(self._targets)
for lib in targets['libraries'] | targets['binaries']:
self._dump_library(lib)
generated_reqs = OrderedSet()
if targets['thrifts']:
for thr in targets['thrifts']:
generated_reqs.add(self._generate_thrift_requirement(thr))
generated_reqs.add(PythonRequirement('thrift', use_2to3=True))
for antlr in targets['antlrs']:
generated_reqs.add(self._generate_antlr_requirement(antlr))
reqs_from_libraries = OrderedSet()
for req_lib in targets['reqs']:
for req in req_lib.payload.requirements:
reqs_from_libraries.add(req)
reqs_to_build = OrderedSet()
find_links = OrderedSet()
for req in reqs_from_libraries | generated_reqs | self._extra_requirements:
if not req.should_build(self._interpreter.python,
Platform.current()):
self.debug('Skipping {} based upon version filter'.format(req))
continue
reqs_to_build.add(req)
self._dump_requirement(req.requirement)
if req.repository:
find_links.add(req.repository)
distributions = self._resolve_multi(reqs_to_build, find_links)
locations = set()
for platform, dist_set in distributions.items():
for dist in dist_set:
if dist.location not in locations:
self._dump_distribution(dist)
locations.add(dist.location)
if len(targets['binaries']) > 1:
print('WARNING: Target has multiple python_binary targets!',
file=sys.stderr)
return self._builder
def _resolve_multi(self, requirements, find_links):
"""Multi-platform dependency resolution for PEX files.
Given a pants configuration and a set of requirements, return a list of distributions
that must be included in order to satisfy them. That may involve distributions for
multiple platforms.
:param requirements: A list of :class:`PythonRequirement` objects to resolve.
:param find_links: Additional paths to search for source packages during resolution.
"""
distributions = dict()
platforms = self.get_platforms(self._platforms
or self._python_setup.platforms)
fetchers = self._python_repos.get_fetchers()
fetchers.extend(Fetcher([path]) for path in find_links)
context = self._python_repos.get_network_context()
for platform in platforms:
requirements_cache_dir = os.path.join(
self._python_setup.resolver_cache_dir,
str(self._interpreter.identity))
distributions[platform] = resolve(
requirements=[req.requirement for req in requirements],
interpreter=self._interpreter,
fetchers=fetchers,
platform=platform,
context=context,
cache=requirements_cache_dir,
cache_ttl=self._python_setup.resolver_cache_ttl)
return distributions
def invalidated(self,
targets,
invalidate_dependents=False,
silent=False,
fingerprint_strategy=None,
topological_order=False):
"""Checks targets for invalidation, first checking the artifact cache.
Subclasses call this to figure out what to work on.
:API: public
:param targets: The targets to check for changes.
:param invalidate_dependents: If True then any targets depending on changed targets are
invalidated.
:param silent: If true, suppress logging information about target invalidation.
:param fingerprint_strategy: A FingerprintStrategy instance, which can do per task,
finer grained fingerprinting of a given Target.
:param topological_order: Whether to invalidate in dependency order.
If no exceptions are thrown by work in the block, the build cache is updated for the targets.
Note: the artifact cache is not updated. That must be done manually.
:returns: Yields an InvalidationCheck object reflecting the targets.
:rtype: InvalidationCheck
"""
cache_key_generator = CacheKeyGenerator(
self.context.options.for_global_scope().cache_key_gen_version,
self.fingerprint)
cache_manager = InvalidationCacheManager(
self.workdir,
cache_key_generator,
self._build_invalidator_dir,
invalidate_dependents,
fingerprint_strategy=fingerprint_strategy,
invalidation_report=self.context.invalidation_report,
task_name=type(self).__name__,
task_version=self.implementation_version_str(),
artifact_write_callback=self.maybe_write_artifact)
invalidation_check = cache_manager.check(
targets, topological_order=topological_order)
self._maybe_create_results_dirs(invalidation_check.all_vts)
if invalidation_check.invalid_vts and self.artifact_cache_reads_enabled(
):
with self.context.new_workunit('cache'):
cached_vts, uncached_vts, uncached_causes = \
self.check_artifact_cache(self.check_artifact_cache_for(invalidation_check))
if cached_vts:
cached_targets = [vt.target for vt in cached_vts]
self.context.run_tracker.artifact_cache_stats.add_hits(
cache_manager.task_name, cached_targets)
if not silent:
self._report_targets('Using cached artifacts for ',
cached_targets, '.')
if uncached_vts:
uncached_targets = [vt.target for vt in uncached_vts]
self.context.run_tracker.artifact_cache_stats.add_misses(
cache_manager.task_name, uncached_targets, uncached_causes)
if not silent:
self._report_targets('No cached artifacts for ',
uncached_targets, '.')
# Now that we've checked the cache, re-partition whatever is still invalid.
invalidation_check = \
InvalidationCheck(invalidation_check.all_vts, uncached_vts)
if not silent:
targets = []
for vt in invalidation_check.invalid_vts:
targets.extend(vt.targets)
if len(targets):
msg_elements = [
'Invalidated ',
items_to_report_element(
[t.address.reference() for t in targets], 'target'),
'.'
]
self.context.log.info(*msg_elements)
invalidation_report = self.context.invalidation_report
if invalidation_report:
for vts in invalidation_check.all_vts:
invalidation_report.add_vts(cache_manager,
vts.targets,
vts.cache_key,
vts.valid,
phase='pre-check')
# Cache has been checked to create the full list of invalid VTs.
# Only copy previous_results for this subset of VTs.
if self.incremental:
for vts in invalidation_check.invalid_vts:
vts.copy_previous_results()
# Yield the result, and then mark the targets as up to date.
yield invalidation_check
if invalidation_report:
for vts in invalidation_check.all_vts:
invalidation_report.add_vts(cache_manager,
vts.targets,
vts.cache_key,
vts.valid,
phase='post-check')
for vt in invalidation_check.invalid_vts:
vt.update()
# Background work to clean up previous builds.
if self.context.options.for_global_scope(
).workdir_max_build_entries is not None:
self._launch_background_workdir_cleanup(invalidation_check.all_vts)
def test_env(content=TEST_CONTENT):
with temporary_dir() as d:
with tempfile.NamedTemporaryFile() as f:
f.write(content)
f.flush()
yield f, CacheKeyGenerator(), BuildInvalidator(d)
def invalidated(self,
targets,
invalidate_dependents=False,
silent=False,
fingerprint_strategy=None,
topological_order=False):
"""Checks targets for invalidation, first checking the artifact cache.
Subclasses call this to figure out what to work on.
:API: public
:param targets: The targets to check for changes.
:param invalidate_dependents: If True then any targets depending on changed targets are
invalidated.
:param silent: If true, suppress logging information about target invalidation.
:param fingerprint_strategy: A FingerprintStrategy instance, which can do per task,
finer grained fingerprinting of a given Target.
:param topological_order: Whether to invalidate in dependency order.
If no exceptions are thrown by work in the block, the build cache is updated for the targets.
Note: the artifact cache is not updated. That must be done manually.
:returns: Yields an InvalidationCheck object reflecting the targets.
:rtype: InvalidationCheck
"""
cache_key_generator = CacheKeyGenerator(
self.context.options.for_global_scope().cache_key_gen_version,
self.fingerprint)
cache_manager = InvalidationCacheManager(
self.workdir,
cache_key_generator,
self._build_invalidator(),
invalidate_dependents,
fingerprint_strategy=fingerprint_strategy,
invalidation_report=self.context.invalidation_report,
task_name=type(self).__name__,
task_version=self.implementation_version_str(),
artifact_write_callback=self.maybe_write_artifact)
# If this Task's execution has been forced, invalidate all our target fingerprints.
if self._cache_factory.ignore and not self._force_invalidated:
self.invalidate()
self._force_invalidated = True
invalidation_check = cache_manager.check(
targets, topological_order=topological_order)
self._maybe_create_results_dirs(invalidation_check.all_vts)
if invalidation_check.invalid_vts and self.artifact_cache_reads_enabled(
):
with self.context.new_workunit('cache'):
cached_vts, uncached_vts, uncached_causes = \
self.check_artifact_cache(self.check_artifact_cache_for(invalidation_check))
if cached_vts:
cached_targets = [vt.target for vt in cached_vts]
self.context.run_tracker.artifact_cache_stats.add_hits(
cache_manager.task_name, cached_targets)
if not silent:
self._report_targets('Using cached artifacts for ',
cached_targets, '.')
if uncached_vts:
uncached_targets = [vt.target for vt in uncached_vts]
self.context.run_tracker.artifact_cache_stats.add_misses(
cache_manager.task_name, uncached_targets, uncached_causes)
if not silent:
self._report_targets('No cached artifacts for ',
uncached_targets, '.')
# Now that we've checked the cache, re-partition whatever is still invalid.
invalidation_check = \
InvalidationCheck(invalidation_check.all_vts, uncached_vts)
if not silent:
targets = []
for vt in invalidation_check.invalid_vts:
targets.extend(vt.targets)
if len(targets):
msg_elements = [
'Invalidated ',
items_to_report_element(
[t.address.reference() for t in targets], 'target'),
'.'
]
self.context.log.info(*msg_elements)
invalidation_report = self.context.invalidation_report
if invalidation_report:
for vts in invalidation_check.all_vts:
invalidation_report.add_vts(cache_manager,
vts.targets,
vts.cache_key,
vts.valid,
phase='pre-check')
# Cache has been checked to create the full list of invalid VTs.
# Only copy previous_results for this subset of VTs.
if self.incremental:
for vts in invalidation_check.invalid_vts:
vts.copy_previous_results()
# This may seem odd: why would we need to invalidate a VersionedTargetSet that is already
# invalid? But the name force_invalidate() is slightly misleading in this context - what it
# actually does is delete the key file created at the end of the last successful task run.
# This is necessary to avoid the following scenario:
#
# 1) In state A: Task suceeds and writes some output. Key is recorded by the invalidator.
# 2) In state B: Task fails, but writes some output. Key is not recorded.
# 3) After reverting back to state A: The current key is the same as the one recorded at the
# end of step 1), so it looks like no work needs to be done, but actually the task
# must re-run, to overwrite the output written in step 2.
#
# Deleting the file ensures that if a task fails, there is no key for which we might think
# we're in a valid state.
for vts in invalidation_check.invalid_vts:
vts.force_invalidate()
# Yield the result, and then mark the targets as up to date.
yield invalidation_check
if invalidation_report:
for vts in invalidation_check.all_vts:
invalidation_report.add_vts(cache_manager,
vts.targets,
vts.cache_key,
vts.valid,
phase='post-check')
for vt in invalidation_check.invalid_vts:
vt.update()
# Background work to clean up previous builds.
if self.context.options.for_global_scope(
).workdir_max_build_entries is not None:
self._launch_background_workdir_cleanup(invalidation_check.all_vts)
