Source code

Revision control

Copy as Markdown

Other Tools

# 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/.
# This module contains code for managing WebIDL files and bindings for
# the build system.
import errno
import hashlib
import io
import json
import logging
import os
import sys
from multiprocessing import Pool
import mozpack.path as mozpath
from mach.mixin.logging import LoggingMixin
from mozbuild.makeutil import Makefile
from mozbuild.pythonutil import iter_modules_in_path
from mozbuild.util import FileAvoidWrite
# There are various imports in this file in functions to avoid adding
# dependencies to config.status. See bug 949875.
# Limit the count on Windows, because of bug 1889842 and also the
# inefficiency of fork on Windows.
DEFAULT_PROCESS_COUNT = 4 if sys.platform == "win32" else os.cpu_count()
class WebIDLPool:
"""
Distribute generation load across several processes, avoiding redundant state
copies.
"""
GeneratorState = None
def __init__(self, GeneratorState, *, processes=None):
if processes is None:
processes = DEFAULT_PROCESS_COUNT
# As a special case, don't spawn an extra process if processes=1
if processes == 1:
WebIDLPool._init(GeneratorState)
class SeqPool:
def map(self, *args):
return list(map(*args))
self.pool = SeqPool()
else:
self.pool = Pool(
initializer=WebIDLPool._init,
initargs=(GeneratorState,),
processes=processes,
)
def run(self, filenames):
return self.pool.map(WebIDLPool._run, filenames)
@staticmethod
def _init(GeneratorState):
WebIDLPool.GeneratorState = GeneratorState
@staticmethod
def _run(filename):
return WebIDLPool.GeneratorState._generate_build_files_for_webidl(filename)
class BuildResult(object):
"""Represents the result of processing WebIDL files.
This holds a summary of output file generation during code generation.
"""
def __init__(self):
# The .webidl files that had their outputs regenerated.
self.inputs = set()
# The output files that were created.
self.created = set()
# The output files that changed.
self.updated = set()
# The output files that didn't change.
self.unchanged = set()
class WebIDLCodegenManagerState(dict):
"""Holds state for the WebIDL code generation manager.
State is currently just an extended dict. The internal implementation of
state should be considered a black box to everyone except
WebIDLCodegenManager. But we'll still document it.
Any set stored in this dict should be copied and sorted in the `dump()`
method.
Fields:
version
The integer version of the format. This is to detect incompatible
changes between state. It should be bumped whenever the format
changes or semantics change.
webidls
A dictionary holding information about every known WebIDL input.
Keys are the basenames of input WebIDL files. Values are dicts of
metadata. Keys in those dicts are:
* filename - The full path to the input filename.
* inputs - A set of full paths to other webidl files this webidl
depends on.
* outputs - Set of full output paths that are created/derived from
this file.
* sha1 - The hexidecimal SHA-1 of the input filename from the last
processing time.
global_inputs
A dictionary defining files that influence all processing. Keys
are full filenames. Values are hexidecimal SHA-1 from the last
processing time.
dictionaries_convertible_to_js
A set of names of dictionaries that are convertible to JS.
dictionaries_convertible_from_js
A set of names of dictionaries that are convertible from JS.
"""
VERSION = 3
def __init__(self, fh=None):
self["version"] = self.VERSION
self["webidls"] = {}
self["global_depends"] = {}
if not fh:
return
state = json.load(fh)
if state["version"] != self.VERSION:
raise Exception("Unknown state version: %s" % state["version"])
self["version"] = state["version"]
self["global_depends"] = state["global_depends"]
for k, v in state["webidls"].items():
self["webidls"][k] = v
# Sets are converted to lists for serialization because JSON
# doesn't support sets.
self["webidls"][k]["inputs"] = set(v["inputs"])
self["webidls"][k]["outputs"] = set(v["outputs"])
self["dictionaries_convertible_to_js"] = set(
state["dictionaries_convertible_to_js"]
)
self["dictionaries_convertible_from_js"] = set(
state["dictionaries_convertible_from_js"]
)
def dump(self, fh):
"""Dump serialized state to a file handle."""
normalized = self.copy()
webidls = normalized["webidls"] = self["webidls"].copy()
for k, v in self["webidls"].items():
webidls_k = webidls[k] = v.copy()
# Convert sets to lists because JSON doesn't support sets.
webidls_k["outputs"] = sorted(v["outputs"])
webidls_k["inputs"] = sorted(v["inputs"])
normalized["dictionaries_convertible_to_js"] = sorted(
self["dictionaries_convertible_to_js"]
)
normalized["dictionaries_convertible_from_js"] = sorted(
self["dictionaries_convertible_from_js"]
)
json.dump(normalized, fh, sort_keys=True)
class WebIDLCodegenManager(LoggingMixin):
"""Manages all code generation around WebIDL.
To facilitate testing, this object is meant to be generic and reusable.
Paths, etc should be parameters and not hardcoded.
"""
# Global parser derived declaration files.
GLOBAL_DECLARE_FILES = {
"BindingNames.h",
"GeneratedAtomList.h",
"GeneratedEventList.h",
"PrototypeList.h",
"RegisterBindings.h",
"RegisterShadowRealmBindings.h",
"RegisterWorkerBindings.h",
"RegisterWorkerDebuggerBindings.h",
"RegisterWorkletBindings.h",
"UnionTypes.h",
"WebIDLPrefs.h",
"WebIDLSerializable.h",
}
# Global parser derived definition files.
GLOBAL_DEFINE_FILES = {
"BindingNames.cpp",
"RegisterBindings.cpp",
"RegisterShadowRealmBindings.cpp",
"RegisterWorkerBindings.cpp",
"RegisterWorkerDebuggerBindings.cpp",
"RegisterWorkletBindings.cpp",
"UnionTypes.cpp",
"PrototypeList.cpp",
"WebIDLPrefs.cpp",
"WebIDLSerializable.cpp",
}
def __init__(
self,
config_path,
webidl_root,
inputs,
exported_header_dir,
codegen_dir,
state_path,
cache_dir=None,
make_deps_path=None,
make_deps_target=None,
):
"""Create an instance that manages WebIDLs in the build system.
config_path refers to a WebIDL config file (e.g. Bindings.conf).
inputs is a 4-tuple describing the input .webidl files and how to
process them. Members are:
(set(.webidl files), set(basenames of exported files),
set(basenames of generated events files),
set(example interface names))
exported_header_dir and codegen_dir are directories where generated
files will be written to.
state_path is the path to a file that will receive JSON state from our
actions.
make_deps_path is the path to a make dependency file that we can
optionally write.
make_deps_target is the target that receives the make dependencies. It
must be defined if using make_deps_path.
"""
self.populate_logger()
input_paths, exported_stems, generated_events_stems, example_interfaces = inputs
self._config_path = config_path
self._webidl_root = webidl_root
self._input_paths = set(input_paths)
self._exported_stems = set(exported_stems)
self._generated_events_stems = set(generated_events_stems)
self._generated_events_stems_as_array = generated_events_stems
self._example_interfaces = set(example_interfaces)
self._exported_header_dir = exported_header_dir
self._codegen_dir = codegen_dir
self._state_path = state_path
self._cache_dir = cache_dir
self._make_deps_path = make_deps_path
self._make_deps_target = make_deps_target
if (make_deps_path and not make_deps_target) or (
not make_deps_path and make_deps_target
):
raise Exception(
"Must define both make_deps_path and make_deps_target "
"if one is defined."
)
self._parser_results = None
self._config = None
self._state = WebIDLCodegenManagerState()
if os.path.exists(state_path):
with io.open(state_path, "r") as fh:
try:
self._state = WebIDLCodegenManagerState(fh=fh)
except Exception as e:
self.log(
logging.WARN,
"webidl_bad_state",
{"msg": str(e)},
"Bad WebIDL state: {msg}",
)
@property
def config(self):
if not self._config:
self._parse_webidl()
return self._config
def generate_build_files(self, *, processes=None):
"""Generate files required for the build.
This function is in charge of generating all the .h/.cpp files derived
from input .webidl files. Please note that there are build actions
required to produce .webidl files and these build actions are
explicitly not captured here: this function assumes all .webidl files
are present and up to date.
This routine is called as part of the build to ensure files that need
to exist are present and up to date. This routine may not be called if
the build dependencies (generated as a result of calling this the first
time) say everything is up to date.
Because reprocessing outputs for every .webidl on every invocation
is expensive, we only regenerate the minimal set of files on every
invocation. The rules for deciding what needs done are roughly as
follows:
1. If any .webidl changes, reparse all .webidl files and regenerate
the global derived files. Only regenerate output files (.h/.cpp)
impacted by the modified .webidl files.
2. If an non-.webidl dependency (Python files, config file) changes,
assume everything is out of date and regenerate the world. This
is because changes in those could globally impact every output
file.
3. If an output file is missing, ensure it is present by performing
necessary regeneration.
if `processes` is set to None, run in parallel using the
multiprocess.Pool default. If set to 1, don't use extra processes.
"""
# Despite #1 above, we assume the build system is smart enough to not
# invoke us if nothing has changed. Therefore, any invocation means
# something has changed. And, if anything has changed, we need to
# parse the WebIDL.
self._parse_webidl()
result = BuildResult()
# If we parse, we always update globals - they are cheap and it is
# easier that way.
created, updated, unchanged = self._write_global_derived()
result.created |= created
result.updated |= updated
result.unchanged |= unchanged
# If any of the extra dependencies changed, regenerate the world.
global_changed, global_hashes = self._global_dependencies_changed()
if global_changed:
# Make a copy because we may modify.
changed_inputs = set(self._input_paths)
else:
changed_inputs = self._compute_changed_inputs()
self._state["global_depends"] = global_hashes
self._state["dictionaries_convertible_to_js"] = set(
d.identifier.name for d in self._config.getDictionariesConvertibleToJS()
)
self._state["dictionaries_convertible_from_js"] = set(
d.identifier.name for d in self._config.getDictionariesConvertibleFromJS()
)
# Distribute the generation load across several processes. This requires
# a) that `self' is serializable and b) that `self' is unchanged by
# _generate_build_files_for_webidl(...)
ordered_changed_inputs = sorted(changed_inputs)
pool = WebIDLPool(self, processes=processes)
generation_results = pool.run(ordered_changed_inputs)
# Generate bindings from .webidl files.
for filename, generation_result in zip(
ordered_changed_inputs, generation_results
):
basename = mozpath.basename(filename)
result.inputs.add(filename)
written, deps = generation_result
result.created |= written[0]
result.updated |= written[1]
result.unchanged |= written[2]
self._state["webidls"][basename] = dict(
filename=filename,
outputs=written[0] | written[1] | written[2],
inputs=set(deps),
sha1=self._input_hashes[filename],
)
# Process some special interfaces required for testing.
for interface in self._example_interfaces:
written = self.generate_example_files(interface)
result.created |= written[0]
result.updated |= written[1]
result.unchanged |= written[2]
# Generate a make dependency file.
if self._make_deps_path:
mk = Makefile()
codegen_rule = mk.create_rule([self._make_deps_target])
codegen_rule.add_dependencies(global_hashes.keys())
codegen_rule.add_dependencies(self._input_paths)
with FileAvoidWrite(self._make_deps_path) as fh:
mk.dump(fh)
self._save_state()
return result
def generate_example_files(self, interface):
"""Generates example files for a given interface."""
from Codegen import CGExampleRoot
root = CGExampleRoot(self.config, interface)
example_paths = self._example_paths(interface)
for path in example_paths:
self.log(
logging.INFO,
"webidl_generate_example_files",
{"filename": path},
"Generating WebIDL example files derived from {filename}",
)
return self._maybe_write_codegen(root, *example_paths)
def _parse_webidl(self):
import WebIDL
from Configuration import Configuration
self.log(
logging.INFO,
"webidl_parse",
{"count": len(self._input_paths)},
"Parsing {count} WebIDL files.",
)
hashes = {}
parser = WebIDL.Parser(self._cache_dir, lexer=None)
for path in sorted(self._input_paths):
with io.open(path, "r", encoding="utf-8") as fh:
data = fh.read()
hashes[path] = hashlib.sha1(data.encode()).hexdigest()
parser.parse(data, path)
# Only these directories may contain WebIDL files with interfaces
# which are exposed to the web. WebIDL files in these roots may not
# be changed without DOM peer review.
#
# Other directories may contain WebIDL files as long as they only
# contain ChromeOnly interfaces. These are not subject to mandatory
# DOM peer review.
web_roots = (
# The main WebIDL root.
self._webidl_root,
# The binding config root, which contains some test-only
# interfaces.
os.path.dirname(self._config_path),
# The objdir sub-directory which contains generated WebIDL files.
self._codegen_dir,
)
self._parser_results = parser.finish()
self._config = Configuration(
self._config_path,
web_roots,
self._parser_results,
self._generated_events_stems_as_array,
)
self._input_hashes = hashes
def _write_global_derived(self):
from Codegen import GlobalGenRoots
things = [("declare", f) for f in self.GLOBAL_DECLARE_FILES]
things.extend(("define", f) for f in self.GLOBAL_DEFINE_FILES)
result = (set(), set(), set())
for what, filename in things:
stem = mozpath.splitext(filename)[0]
root = getattr(GlobalGenRoots, stem)(self._config)
if what == "declare":
code = root.declare()
output_root = self._exported_header_dir
elif what == "define":
code = root.define()
output_root = self._codegen_dir
else:
raise Exception("Unknown global gen type: %s" % what)
output_path = mozpath.join(output_root, filename)
self._maybe_write_file(output_path, code, result)
return result
def _compute_changed_inputs(self):
"""Compute the set of input files that need to be regenerated."""
changed_inputs = set()
expected_outputs = self.expected_build_output_files()
# Look for missing output files.
if any(not os.path.exists(f) for f in expected_outputs):
# FUTURE Bug 940469 Only regenerate minimum set.
changed_inputs |= self._input_paths
# That's it for examining output files. We /could/ examine SHA-1's of
# output files from a previous run to detect modifications. But that's
# a lot of extra work and most build systems don't do that anyway.
# Now we move on to the input files.
old_hashes = {v["filename"]: v["sha1"] for v in self._state["webidls"].values()}
old_filenames = set(old_hashes.keys())
new_filenames = self._input_paths
# If an old file has disappeared or a new file has arrived, mark
# it.
changed_inputs |= old_filenames ^ new_filenames
# For the files in common between runs, compare content. If the file
# has changed, mark it. We don't need to perform mtime comparisons
# because content is a stronger validator.
for filename in old_filenames & new_filenames:
if old_hashes[filename] != self._input_hashes[filename]:
changed_inputs.add(filename)
# We've now populated the base set of inputs that have changed.
# Inherit dependencies from previous run. The full set of dependencies
# is associated with each record, so we don't need to perform any fancy
# graph traversal.
for v in self._state["webidls"].values():
if any(dep for dep in v["inputs"] if dep in changed_inputs):
changed_inputs.add(v["filename"])
# Now check for changes to the set of dictionaries that are convertible to JS
oldDictionariesConvertibleToJS = self._state["dictionaries_convertible_to_js"]
newDictionariesConvertibleToJS = self._config.getDictionariesConvertibleToJS()
newNames = set(d.identifier.name for d in newDictionariesConvertibleToJS)
changedDictionaryNames = oldDictionariesConvertibleToJS ^ newNames
# Now check for changes to the set of dictionaries that are convertible from JS
oldDictionariesConvertibleFromJS = self._state[
"dictionaries_convertible_from_js"
]
newDictionariesConvertibleFromJS = (
self._config.getDictionariesConvertibleFromJS()
)
newNames = set(d.identifier.name for d in newDictionariesConvertibleFromJS)
changedDictionaryNames |= oldDictionariesConvertibleFromJS ^ newNames
for name in changedDictionaryNames:
d = self._config.getDictionaryIfExists(name)
if d:
changed_inputs.add(d.filename)
# Only use paths that are known to our current state.
# This filters out files that were deleted or changed type (e.g. from
# static to preprocessed).
return changed_inputs & self._input_paths
def _binding_info(self, p):
"""Compute binding metadata for an input path.
Returns a tuple of:
(stem, binding_stem, is_event, output_files)
output_files is itself a tuple. The first two items are the binding
header and C++ paths, respectively. The 2nd pair are the event header
and C++ paths or None if this isn't an event binding.
"""
basename = mozpath.basename(p)
stem = mozpath.splitext(basename)[0]
binding_stem = "%sBinding" % stem
if stem in self._exported_stems:
header_dir = self._exported_header_dir
else:
header_dir = self._codegen_dir
is_event = stem in self._generated_events_stems
files = (
mozpath.join(header_dir, "%s.h" % binding_stem),
mozpath.join(self._codegen_dir, "%s.cpp" % binding_stem),
mozpath.join(header_dir, "%s.h" % stem) if is_event else None,
mozpath.join(self._codegen_dir, "%s.cpp" % stem) if is_event else None,
)
return stem, binding_stem, is_event, header_dir, files
def _example_paths(self, interface):
return (
mozpath.join(self._codegen_dir, "%s-example.h" % interface),
mozpath.join(self._codegen_dir, "%s-example.cpp" % interface),
)
def expected_build_output_files(self):
"""Obtain the set of files generate_build_files() should write."""
paths = set()
# Account for global generation.
for p in self.GLOBAL_DECLARE_FILES:
paths.add(mozpath.join(self._exported_header_dir, p))
for p in self.GLOBAL_DEFINE_FILES:
paths.add(mozpath.join(self._codegen_dir, p))
for p in self._input_paths:
stem, binding_stem, is_event, header_dir, files = self._binding_info(p)
paths |= {f for f in files if f}
for interface in self._example_interfaces:
for p in self._example_paths(interface):
paths.add(p)
return paths
# Parallelization of the generation step relies on this method not changing
# the internal state of the object
def _generate_build_files_for_webidl(self, filename):
from Codegen import CGBindingRoot, CGEventRoot
self.log(
logging.INFO,
"webidl_generate_build_for_input",
{"filename": filename},
"Generating WebIDL files derived from {filename}",
)
stem, binding_stem, is_event, header_dir, files = self._binding_info(filename)
root = CGBindingRoot(self._config, binding_stem, filename)
result = self._maybe_write_codegen(root, files[0], files[1])
if is_event:
generated_event = CGEventRoot(self._config, stem)
result = self._maybe_write_codegen(
generated_event, files[2], files[3], result
)
return result, root.deps()
def _global_dependencies_changed(self):
"""Determine whether the global dependencies have changed."""
current_files = set(iter_modules_in_path(mozpath.dirname(__file__)))
# We need to catch other .py files from /dom/bindings. We assume these
# are in the same directory as the config file.
current_files |= set(iter_modules_in_path(mozpath.dirname(self._config_path)))
current_files.add(self._config_path)
current_hashes = {}
for f in current_files:
# This will fail if the file doesn't exist. If a current global
# dependency doesn't exist, something else is wrong.
with io.open(f, "rb") as fh:
current_hashes[f] = hashlib.sha1(fh.read()).hexdigest()
# The set of files has changed.
if current_files ^ set(self._state["global_depends"].keys()):
return True, current_hashes
# Compare hashes.
for f, sha1 in current_hashes.items():
if sha1 != self._state["global_depends"][f]:
return True, current_hashes
return False, current_hashes
def _save_state(self):
with io.open(self._state_path, "w", newline="\n") as fh:
self._state.dump(fh)
def _maybe_write_codegen(self, obj, declare_path, define_path, result=None):
assert declare_path and define_path
if not result:
result = (set(), set(), set())
self._maybe_write_file(declare_path, obj.declare(), result)
self._maybe_write_file(define_path, obj.define(), result)
return result
def _maybe_write_file(self, path, content, result):
fh = FileAvoidWrite(path)
fh.write(content)
existed, updated = fh.close()
if not existed:
result[0].add(path)
elif updated:
result[1].add(path)
else:
result[2].add(path)
def create_build_system_manager(topsrcdir=None, topobjdir=None, dist_dir=None):
"""Create a WebIDLCodegenManager for use by the build system."""
if topsrcdir is None:
assert topobjdir is None and dist_dir is None
import buildconfig
topsrcdir = buildconfig.topsrcdir
topobjdir = buildconfig.topobjdir
dist_dir = buildconfig.substs["DIST"]
src_dir = os.path.join(topsrcdir, "dom", "bindings")
obj_dir = os.path.join(topobjdir, "dom", "bindings")
webidl_root = os.path.join(topsrcdir, "dom", "webidl")
with io.open(os.path.join(obj_dir, "file-lists.json"), "r") as fh:
files = json.load(fh)
inputs = (
files["webidls"],
files["exported_stems"],
files["generated_events_stems"],
files["example_interfaces"],
)
cache_dir = os.path.join(obj_dir, "_cache")
try:
os.makedirs(cache_dir)
except OSError as e:
if e.errno != errno.EEXIST:
raise
return WebIDLCodegenManager(
os.path.join(src_dir, "Bindings.conf"),
webidl_root,
inputs,
os.path.join(dist_dir, "include", "mozilla", "dom"),
obj_dir,
os.path.join(obj_dir, "codegen.json"),
cache_dir=cache_dir,
# The make rules include a codegen.pp file containing dependencies.
make_deps_path=os.path.join(obj_dir, "codegen.pp"),
make_deps_target="webidl.stub",
)