corysanin/tubestation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Bug 1759030 - [gecko_taskgraph] Use 'generator.py' from upstream Taskgraph, r=releng-reviewers,jlorenzo

Browse Source 
Differential Revision: https://phabricator.services.mozilla.com/D161057
This commit is contained in:
Andrew Halberstadt
2022-11-04 14:14:57 +00:00
parent 08e5b0e5fa
commit 599e696d24
12 changed files with 14 additions and 601 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
python/mozbuild/mozbuild/toolchains.py
View File

@@ -8,7 +8,8 @@ import six
def toolchain_task_definitions():
from gecko_taskgraph.generator import load_tasks_for_kind
import gecko_taskgraph # noqa: triggers override of the `graph_config_schema`
from taskgraph.generator import load_tasks_for_kind
# Don't import globally to allow this module being imported without
# the taskgraph module being available (e.g. standalone js)

2
taskcluster/gecko_taskgraph/__init__.py
View File

@@ -46,11 +46,11 @@ def register(graph_config):
Args:
graph_config: The graph configuration object.
"""
from taskgraph import generator
from gecko_taskgraph.parameters import register_parameters
from gecko_taskgraph import ( # noqa: trigger target task method registration
target_tasks,
)
from gecko_taskgraph import generator
from gecko_taskgraph import morph # noqa: trigger morph registration
from gecko_taskgraph.util.verify import verifications

2
taskcluster/gecko_taskgraph/decision.py
View File

@@ -22,6 +22,7 @@ from taskgraph.decision import (
_determine_more_accurate_base_rev,
_get_env_prefix,
)
from taskgraph.generator import TaskGraphGenerator
from taskgraph.parameters import Parameters
from taskgraph.taskgraph import TaskGraph
from taskgraph.util.python_path import find_object
@@ -33,7 +34,6 @@ from voluptuous import Any, Optional, Required
from . import GECKO
from .actions import render_actions_json
from .generator import TaskGraphGenerator
from .parameters import get_app_version, get_version
from .try_option_syntax import parse_message
from .util.backstop import BACKSTOP_INDEX, is_backstop

2
taskcluster/gecko_taskgraph/docker.py
View File

@@ -8,10 +8,10 @@ import os
import tarfile
from io import BytesIO
from taskgraph.generator import load_tasks_for_kind
from taskgraph.parameters import Parameters
from taskgraph.util.taskcluster import get_session, get_artifact_url
from gecko_taskgraph.generator import load_tasks_for_kind
from gecko_taskgraph.optimize.strategies import IndexSearch
from gecko_taskgraph.util import docker
from . import GECKO

447
taskcluster/gecko_taskgraph/generator.py
View File

@@ -1,447 +0,0 @@
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
import logging
import os
import copy
import attr
from taskgraph import filter_tasks
from taskgraph.config import GraphConfig, load_graph_config
from taskgraph.graph import Graph
from taskgraph.morph import morph
from taskgraph.optimize.base import optimize_task_graph
from taskgraph.parameters import parameters_loader
from taskgraph.task import Task
from taskgraph.taskgraph import TaskGraph
from taskgraph.transforms.base import TransformSequence, TransformConfig
from taskgraph.util.python_path import find_object
from taskgraph.util.yaml import load_yaml
from taskgraph.util.verify import verifications
logger = logging.getLogger(__name__)
class KindNotFound(Exception):
"""
Raised when trying to load kind from a directory without a kind.yml.
"""
@attr.s(frozen=True)
class Kind:
name = attr.ib(type=str)
path = attr.ib(type=str)
config = attr.ib(type=dict)
graph_config = attr.ib(type=GraphConfig)
def _get_loader(self):
try:
loader = self.config["loader"]
except KeyError:
raise KeyError(f"{self.path!r} does not define `loader`")
return find_object(loader)
def load_tasks(self, parameters, loaded_tasks, write_artifacts):
loader = self._get_loader()
config = copy.deepcopy(self.config)
kind_dependencies = config.get("kind-dependencies", [])
kind_dependencies_tasks = {
task.label: task for task in loaded_tasks if task.kind in kind_dependencies
}
inputs = loader(self.name, self.path, config, parameters, loaded_tasks)
transforms = TransformSequence()
for xform_path in config["transforms"]:
transform = find_object(xform_path)
transforms.add(transform)
# perform the transformations on the loaded inputs
trans_config = TransformConfig(
self.name,
self.path,
config,
parameters,
kind_dependencies_tasks,
self.graph_config,
write_artifacts=write_artifacts,
)
tasks = [
Task(
self.name,
label=task_dict["label"],
description=task_dict["description"],
attributes=task_dict["attributes"],
task=task_dict["task"],
optimization=task_dict.get("optimization"),
dependencies=task_dict.get("dependencies"),
soft_dependencies=task_dict.get("soft-dependencies"),
if_dependencies=task_dict.get("if-dependencies"),
)
for task_dict in transforms(trans_config, inputs)
]
return tasks
@classmethod
def load(cls, root_dir, graph_config, kind_name):
path = os.path.join(root_dir, kind_name)
kind_yml = os.path.join(path, "kind.yml")
if not os.path.exists(kind_yml):
raise KindNotFound(kind_yml)
logger.debug(f"loading kind `{kind_name}` from `{path}`")
config = load_yaml(kind_yml)
return cls(kind_name, path, config, graph_config)
class TaskGraphGenerator:
"""
The central controller for taskgraph. This handles all phases of graph
generation. The task is generated from all of the kinds defined in
subdirectories of the generator's root directory.
Access to the results of this generation, as well as intermediate values at
various phases of generation, is available via properties. This encourages
the provision of all generation inputs at instance construction time.
"""
# Task-graph generation is implemented as a Python generator that yields
# each "phase" of generation. This allows some mach subcommands to short-
# circuit generation of the entire graph by never completing the generator.
def __init__(
self,
root_dir,
parameters,
decision_task_id="DECISION-TASK",
write_artifacts=False,
):
"""
@param root_dir: root directory, with subdirectories for each kind
@param paramaters: parameters for this task-graph generation, or callable
taking a `GraphConfig` and returning parameters
@type parameters: Union[Parameters, Callable[[GraphConfig], Parameters]]
"""
if root_dir is None:
root_dir = "taskcluster/ci"
self.root_dir = root_dir
self._parameters = parameters
self._decision_task_id = decision_task_id
self._write_artifacts = write_artifacts
# start the generator
self._run = self._run()
self._run_results = {}
@property
def parameters(self):
"""
The properties used for this graph.
@type: Properties
"""
return self._run_until("parameters")
@property
def full_task_set(self):
"""
The full task set: all tasks defined by any kind (a graph without edges)
@type: TaskGraph
"""
return self._run_until("full_task_set")
@property
def full_task_graph(self):
"""
The full task graph: the full task set, with edges representing
dependencies.
@type: TaskGraph
"""
return self._run_until("full_task_graph")
@property
def target_task_set(self):
"""
The set of targetted tasks (a graph without edges)
@type: TaskGraph
"""
return self._run_until("target_task_set")
@property
def target_task_graph(self):
"""
The set of targetted tasks and all of their dependencies
@type: TaskGraph
"""
return self._run_until("target_task_graph")
@property
def optimized_task_graph(self):
"""
The set of targetted tasks and all of their dependencies; tasks that
have been optimized out are either omitted or replaced with a Task
instance containing only a task_id.
@type: TaskGraph
"""
return self._run_until("optimized_task_graph")
@property
def label_to_taskid(self):
"""
A dictionary mapping task label to assigned taskId. This property helps
in interpreting `optimized_task_graph`.
@type: dictionary
"""
return self._run_until("label_to_taskid")
@property
def morphed_task_graph(self):
"""
The optimized task graph, with any subsequent morphs applied. This graph
will have the same meaning as the optimized task graph, but be in a form
more palatable to TaskCluster.
@type: TaskGraph
"""
return self._run_until("morphed_task_graph")
@property
def graph_config(self):
"""
The configuration for this graph.
@type: TaskGraph
"""
return self._run_until("graph_config")
def _load_kinds(self, graph_config, target_kind=None):
if target_kind:
# docker-image is an implicit dependency that never appears in
# kind-dependencies.
queue = [target_kind, "docker-image"]
seen_kinds = set()
while queue:
kind_name = queue.pop()
if kind_name in seen_kinds:
continue
seen_kinds.add(kind_name)
kind = Kind.load(self.root_dir, graph_config, kind_name)
yield kind
queue.extend(kind.config.get("kind-dependencies", []))
else:
for kind_name in os.listdir(self.root_dir):
try:
yield Kind.load(self.root_dir, graph_config, kind_name)
except KindNotFound:
continue
def _run(self):
logger.info("Loading graph configuration.")
graph_config = load_graph_config(self.root_dir)
yield ("graph_config", graph_config)
graph_config.register()
# Initial verifications that don't depend on any generation state.
verifications("initial")
if callable(self._parameters):
parameters = self._parameters(graph_config)
else:
parameters = self._parameters
logger.info("Using {}".format(parameters))
logger.debug("Dumping parameters:\n{}".format(repr(parameters)))
filters = parameters.get("filters", [])
# Always add legacy target tasks method until we deprecate that API.
if "target_tasks_method" not in filters:
filters.insert(0, "target_tasks_method")
filters = [filter_tasks.filter_task_functions[f] for f in filters]
yield self.verify("parameters", parameters)
logger.info("Loading kinds")
# put the kinds into a graph and sort topologically so that kinds are loaded
# in post-order
if parameters.get("target-kind"):
target_kind = parameters["target-kind"]
logger.info(
"Limiting kinds to {target_kind} and dependencies".format(
target_kind=target_kind
)
)
kinds = {
kind.name: kind
for kind in self._load_kinds(graph_config, parameters.get("target-kind"))
}
verifications("kinds", kinds)
edges = set()
for kind in kinds.values():
for dep in kind.config.get("kind-dependencies", []):
edges.add((kind.name, dep, "kind-dependency"))
kind_graph = Graph(set(kinds), edges)
if parameters.get("target-kind"):
kind_graph = kind_graph.transitive_closure({target_kind, "docker-image"})
logger.info("Generating full task set")
all_tasks = {}
for kind_name in kind_graph.visit_postorder():
logger.debug(f"Loading tasks for kind {kind_name}")
kind = kinds[kind_name]
try:
new_tasks = kind.load_tasks(
parameters,
list(all_tasks.values()),
self._write_artifacts,
)
except Exception:
logger.exception(f"Error loading tasks for kind {kind_name}:")
raise
for task in new_tasks:
if task.label in all_tasks:
raise Exception("duplicate tasks with label " + task.label)
all_tasks[task.label] = task
logger.info(f"Generated {len(new_tasks)} tasks for kind {kind_name}")
full_task_set = TaskGraph(all_tasks, Graph(set(all_tasks), set()))
yield self.verify("full_task_set", full_task_set, graph_config, parameters)
logger.info("Generating full task graph")
edges = set()
for t in full_task_set:
for depname, dep in t.dependencies.items():
edges.add((t.label, dep, depname))
full_task_graph = TaskGraph(all_tasks, Graph(full_task_set.graph.nodes, edges))
logger.info(
"Full task graph contains %d tasks and %d dependencies"
% (len(full_task_set.graph.nodes), len(edges))
)
yield self.verify("full_task_graph", full_task_graph, graph_config, parameters)
logger.info("Generating target task set")
target_task_set = TaskGraph(
dict(all_tasks), Graph(set(all_tasks.keys()), set())
)
for fltr in filters:
old_len = len(target_task_set.graph.nodes)
target_tasks = set(fltr(target_task_set, parameters, graph_config))
target_task_set = TaskGraph(
{l: all_tasks[l] for l in target_tasks}, Graph(target_tasks, set())
)
logger.info(
"Filter %s pruned %d tasks (%d remain)"
% (fltr.__name__, old_len - len(target_tasks), len(target_tasks))
)
yield self.verify("target_task_set", target_task_set, graph_config, parameters)
logger.info("Generating target task graph")
# include all docker-image build tasks here, in case they are needed for a graph morph
docker_image_tasks = {
t.label
for t in full_task_graph.tasks.values()
if t.attributes["kind"] == "docker-image"
}
# include all tasks with `always_target` set
if parameters["enable_always_target"]:
always_target_tasks = {
t.label
for t in full_task_graph.tasks.values()
if t.attributes.get("always_target")
}
else:
always_target_tasks = set()
logger.info(
"Adding %d tasks with `always_target` attribute"
% (len(always_target_tasks) - len(always_target_tasks & target_tasks))
)
requested_tasks = target_tasks | docker_image_tasks | always_target_tasks
target_graph = full_task_graph.graph.transitive_closure(requested_tasks)
target_task_graph = TaskGraph(
{l: all_tasks[l] for l in target_graph.nodes}, target_graph
)
yield self.verify(
"target_task_graph", target_task_graph, graph_config, parameters
)
logger.info("Generating optimized task graph")
existing_tasks = parameters.get("existing_tasks")
do_not_optimize = set(parameters.get("do_not_optimize", []))
if not parameters.get("optimize_target_tasks", True):
do_not_optimize = set(target_task_set.graph.nodes).union(do_not_optimize)
# this is used for testing experimental optimization strategies
strategies = os.environ.get(
"TASKGRAPH_OPTIMIZE_STRATEGIES", parameters.get("optimize_strategies")
)
if strategies:
strategies = find_object(strategies)
optimized_task_graph, label_to_taskid = optimize_task_graph(
target_task_graph,
requested_tasks,
parameters,
do_not_optimize,
self._decision_task_id,
existing_tasks=existing_tasks,
strategy_override=strategies,
)
yield self.verify(
"optimized_task_graph", optimized_task_graph, graph_config, parameters
)
morphed_task_graph, label_to_taskid = morph(
optimized_task_graph, label_to_taskid, parameters, graph_config
)
yield "label_to_taskid", label_to_taskid
yield self.verify(
"morphed_task_graph", morphed_task_graph, graph_config, parameters
)
def _run_until(self, name):
while name not in self._run_results:
try:
k, v = next(self._run)
except StopIteration:
raise AttributeError(f"No such run result {name}")
self._run_results[k] = v
return self._run_results[name]
def verify(self, name, obj, *args, **kwargs):
verifications(name, obj, *args, **kwargs)
return name, obj
def load_tasks_for_kind(parameters, kind, root_dir=None):
"""
Get all the tasks of a given kind.
This function is designed to be called from outside of taskgraph.
"""
# make parameters read-write
parameters = dict(parameters)
parameters["target-kind"] = kind
parameters = parameters_loader(spec=None, strict=False, overrides=parameters)
tgg = TaskGraphGenerator(root_dir=root_dir, parameters=parameters)
return {
task.task["metadata"]["name"]: task
for task in tgg.full_task_set
if task.kind == kind
}

2
taskcluster/gecko_taskgraph/main.py
View File

@@ -101,7 +101,7 @@ FORMAT_METHODS = {
def get_taskgraph_generator(root, parameters):
"""Helper function to make testing a little easier."""
from gecko_taskgraph.generator import TaskGraphGenerator
from taskgraph.generator import TaskGraphGenerator
return TaskGraphGenerator(root_dir=root, parameters=parameters)

11
taskcluster/gecko_taskgraph/test/conftest.py
View File

@@ -6,18 +6,15 @@ import os
import pytest
from mach.logging import LoggingManager
from responses import RequestsMock
from taskgraph import target_tasks as target_tasks_mod
from taskgraph import generator as generator_mod, target_tasks as target_tasks_mod
from taskgraph.config import GraphConfig, load_graph_config
from taskgraph.generator import TaskGraphGenerator, Kind
from taskgraph.optimize import base as optimize_mod
from taskgraph.optimize.base import OptimizationStrategy
from taskgraph.parameters import Parameters
from gecko_taskgraph import (
GECKO,
generator,
)
from gecko_taskgraph import GECKO
from gecko_taskgraph.actions import render_actions_json
from gecko_taskgraph.generator import TaskGraphGenerator, Kind
from gecko_taskgraph.util.templates import merge
@@ -169,7 +166,7 @@ def maketgg(monkeypatch):
)
parameters.update(params)
monkeypatch.setattr(generator, "load_graph_config", fake_load_graph_config)
monkeypatch.setattr(generator_mod, "load_graph_config", fake_load_graph_config)
tgg = WithFakeKind("/root", parameters)
tgg.loaded_kinds = loaded_kinds

1
taskcluster/gecko_taskgraph/test/python.ini
View File

@@ -4,7 +4,6 @@ subsuite = taskgraph
[test_actions_util.py]
[test_decision.py]
[test_files_changed.py]
[test_generator.py]
[test_main.py]
[test_morph.py]
[test_optimize_strategies.py]

137
taskcluster/gecko_taskgraph/test/test_generator.py
View File

@@ -1,137 +0,0 @@
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
from mozunit import main
from taskgraph.graph import Graph
from gecko_taskgraph.generator import load_tasks_for_kind
from gecko_taskgraph import (
generator,
)
from conftest import (
FakeKind,
WithFakeKind,
fake_load_graph_config,
)
def test_kind_ordering(maketgg):
"When task kinds depend on each other, they are loaded in postorder"
tgg = maketgg(
kinds=[
("_fake3", {"kind-dependencies": ["_fake2", "_fake1"]}),
("_fake2", {"kind-dependencies": ["_fake1"]}),
("_fake1", {"kind-dependencies": []}),
]
)
tgg._run_until("full_task_set")
assert tgg.loaded_kinds == ["_fake1", "_fake2", "_fake3"]
def test_full_task_set(maketgg):
"The full_task_set property has all tasks"
tgg = maketgg()
assert tgg.full_task_set.graph == Graph(
{"_fake-t-0", "_fake-t-1", "_fake-t-2"}, set()
)
assert sorted(tgg.full_task_set.tasks.keys()) == sorted(
["_fake-t-0", "_fake-t-1", "_fake-t-2"]
)
def test_full_task_graph(maketgg):
"The full_task_graph property has all tasks, and links"
tgg = maketgg()
assert tgg.full_task_graph.graph == Graph(
{"_fake-t-0", "_fake-t-1", "_fake-t-2"},
{
("_fake-t-1", "_fake-t-0", "prev"),
("_fake-t-2", "_fake-t-1", "prev"),
},
)
assert sorted(tgg.full_task_graph.tasks.keys()) == sorted(
["_fake-t-0", "_fake-t-1", "_fake-t-2"]
)
def test_target_task_set(maketgg):
"The target_task_set property has the targeted tasks"
tgg = maketgg(["_fake-t-1"])
assert tgg.target_task_set.graph == Graph({"_fake-t-1"}, set())
assert set(tgg.target_task_set.tasks.keys()) == {"_fake-t-1"}
def test_target_task_graph(maketgg):
"The target_task_graph property has the targeted tasks and deps"
tgg = maketgg(["_fake-t-1"])
assert tgg.target_task_graph.graph == Graph(
{"_fake-t-0", "_fake-t-1"}, {("_fake-t-1", "_fake-t-0", "prev")}
)
assert sorted(tgg.target_task_graph.tasks.keys()) == sorted(
["_fake-t-0", "_fake-t-1"]
)
def test_always_target_tasks(maketgg):
"The target_task_graph includes tasks with 'always_target'"
tgg_args = {
"target_tasks": ["_fake-t-0", "_fake-t-1", "_ignore-t-0", "_ignore-t-1"],
"kinds": [
("_fake", {"job-defaults": {"optimization": {"odd": None}}}),
(
"_ignore",
{
"job-defaults": {
"attributes": {"always_target": True},
"optimization": {"even": None},
}
},
),
],
"params": {"optimize_target_tasks": False, "enable_always_target": True},
}
tgg = maketgg(**tgg_args)
assert sorted(tgg.target_task_set.tasks.keys()) == sorted(
["_fake-t-0", "_fake-t-1", "_ignore-t-0", "_ignore-t-1"]
)
assert sorted(tgg.target_task_graph.tasks.keys()) == sorted(
["_fake-t-0", "_fake-t-1", "_ignore-t-0", "_ignore-t-1", "_ignore-t-2"]
)
assert sorted(t.label for t in tgg.optimized_task_graph.tasks.values()) == sorted(
["_fake-t-0", "_fake-t-1", "_ignore-t-0", "_ignore-t-1"]
)
def test_optimized_task_graph(maketgg):
"The optimized task graph contains task ids"
tgg = maketgg(["_fake-t-2"])
tid = tgg.label_to_taskid
assert tgg.optimized_task_graph.graph == Graph(
{tid["_fake-t-0"], tid["_fake-t-1"], tid["_fake-t-2"]},
{
(tid["_fake-t-1"], tid["_fake-t-0"], "prev"),
(tid["_fake-t-2"], tid["_fake-t-1"], "prev"),
},
)
def test_load_tasks_for_kind(monkeypatch):
"""
`load_tasks_for_kinds` will load the tasks for the provided kind
"""
FakeKind.loaded_kinds = []
monkeypatch.setattr(generator, "TaskGraphGenerator", WithFakeKind)
monkeypatch.setattr(generator, "load_graph_config", fake_load_graph_config)
tasks = load_tasks_for_kind(
{"_kinds": [("_example-kind", []), ("docker-image", [])]},
"_example-kind",
"/root",
)
assert "t-1" in tasks and tasks["t-1"].label == "_example-kind-t-1"
if __name__ == "__main__":
main()

4
taskcluster/mach_commands.py
View File

@@ -331,7 +331,7 @@ def setup_logging(command_context, quiet=False, verbose=True):
def show_actions(command_context, options):
import gecko_taskgraph
import gecko_taskgraph.actions
import gecko_taskgraph.generator
from taskgraph.generator import TaskGraphGenerator
from taskgraph.parameters import parameters_loader
try:
@@ -340,7 +340,7 @@ def show_actions(command_context, options):
)
parameters = parameters_loader(options["parameters"])
tgg = gecko_taskgraph.generator.TaskGraphGenerator(
tgg = TaskGraphGenerator(
root_dir=options.get("root"),
parameters=parameters,
)

2
taskcluster/test/conftest.py
View File

@@ -9,9 +9,9 @@ import os
import pytest
from responses import RequestsMock, logger as rsps_logger
from taskgraph.generator import TaskGraphGenerator
from taskgraph.parameters import parameters_loader
from gecko_taskgraph.generator import TaskGraphGenerator
from gecko_taskgraph.util.hg import PUSHLOG_PUSHES_TMPL
from gecko_taskgraph.util.bugbug import BUGBUG_BASE_URL

2
tools/tryselect/tasks.py
View File

@@ -15,10 +15,10 @@ from mach.util import get_state_dir
from mozbuild.base import MozbuildObject
from mozpack.files import FileFinder
from moztest.resolve import TestResolver, TestManifestLoader, get_suite_definition
from taskgraph.generator import TaskGraphGenerator
from taskgraph.parameters import ParameterMismatch, parameters_loader
from taskgraph.taskgraph import TaskGraph
from gecko_taskgraph.generator import TaskGraphGenerator
here = os.path.abspath(os.path.dirname(__file__))
build = MozbuildObject.from_environment(cwd=here)